# Christian T. Macion — Portfolio (llms-full.txt) > Concatenated Markdown export of every public page on the portfolio. This file > is a complement to [/llms.txt](/llms.txt) — the short index. Agents that want > the full text content should read this file. Each section is one page, in > canonical reading order, separated by `---`. > Build: 2026-08-08T19:06:57Z. Regenerated on every deploy. Generated by `scripts/build-llms-full.mjs`. --- # Christian T. Macion · Quantitative Researcher + AI Engineer-Architect for financial markets > Christian T. Macion. Quantitative Researcher and AI Engineer-Architect with 6 years across quant research, production trading systems, and multi-agent AI for financial markets. Equities, futures, FX, crypto, and alt-data. 76.5k LOC Python, 25 public repositories, and 102 professional certifications. NDA-safe by construction. Source: `/` _(Christian T. Macion, professional portrait, three-quarter view. Photographed January 2026.)_ CHRISTIAN T. MACION · DIGOS CITY, DAVAO DEL SUR, PHILIPPINES · REMOTE (UTC+8) # Quantitative Researcher AI Engineer-Architect. 6 years across quant research, production trading systems, and multi-agent AI for financial markets. I'm a quant researcher who builds the eval harness behind the alpha. Six years across equities, futures, FX, and crypto. Same methodology, two lanes: systematic strategies on public data and multi-agent LLM systems behind frozen specs. I love turning rough ideas into things a stranger can re-run from a public commit. The proof sits one click below: strategy workbooks, trade logs, and the G1 to G31 gate stack. [github ↗](https://github.com/christianmacion)[linkedin ↗](https://www.linkedin.com/in/christianmacion)[medium ↗](https://medium.com/@christianmacion)[email ↗](mailto:christianmacion26@gmail.com) FINANCIAL MARKETS · EQUITIES · FUTURES · FX · CRYPTO · ALT-DATA·[1-PAGE RESUME ↗](/portfolio/resume-quant-onepage.pdf) 00. live performance PAPER-TRADE LEDGER · EXPECTED · NOT HISTORICAL +60%+40%+20%0%M0M5M10M15NOW+47.3%↗ 6 YEARS EXPERIENCE ↗ 9 QUANT STRATEGIES ↗ 76.5k PYTHON LOC ↗ 6 AI PROJECTS ↗ 25 PUBLIC GITHUB REPOS ↗ 102 CERTIFICATIONS 102 certs in the last 11 months. AWS, GCP, quant, AI. 6 YEARS · QUANT RESEARCH + AI ENGINEERING9 PAPER-TRADED ALPHA STRATEGIES76.5K PYTHON LOC6 AI ENGINEERING PROJECTS SHIPPED25 PUBLIC GITHUB REPOSITORIESSTA TIER-1 CERTIFIED TECHNICAL ANALYST6 YEARS · QUANT RESEARCH + AI ENGINEERING9 PAPER-TRADED ALPHA STRATEGIES76.5K PYTHON LOC6 AI ENGINEERING PROJECTS SHIPPED25 PUBLIC GITHUB REPOSITORIESSTA TIER-1 CERTIFIED TECHNICAL ANALYSTframe loop 01 / projects ## § 01. The Work Quant research, production AI engineering, and the chrome discipline that makes the work legible to a hiring reader are not adjacent skills. They are one cross-validated practice. A trading edge cleared by the 31-gate stack is the same artifact audited for accessibility, indexed in a workbook, and rendered as recruiter-facing proof on this site. Each discipline stress-tests the others. Below: 9 quant strategies with published methodology, 6 production AI projects, 25 public repositories, and a 31-gate eval harness with a public kill log. Every claim is dated. Every commit is referenced. Every closed backtest carries its drawdown window. Client-confidential work is omitted. AI · HISTORY · THE LEAD ## Always up to trend. Two months ahead. Six architectural jumps in three years, each adopted on day one. The lead is the trend-tracking fingerprint for a quant seat. 11MONTHS AHEAD·7STAGES·2022 to 2025ARC·112AGENTS NOW PROMPTS Nov 2022→Dec 2022 I adopted Dec 2022 1months ahead Single-turn Q&A; research summarization SKILLS Jul 2023→Aug 2023 I adopted Aug 2023 1months ahead Tools/functions; structured output AGENTS Dec 2023→Feb 2024 I adopted Feb 2024 2months ahead Single-agent loops; tasks HARNESS Apr 2024→Jun 2024 I adopted Jun 2024 2months ahead AI writes code in loop LOOPS Sep 2024→Nov 2024 I adopted Nov 2024 2months ahead Persistent agentic workflows GRAPHS Mar 2025→May 2025 I adopted May 2025 2months ahead Persistent state + memory; knowledge graphs OFFICES Q4 2025→Q4 2025 I adopted Q4 2025 1months ahead 14 squads · graph-engineered multi-agent orchestration NOW 02 / proof ## § 02. What I Ship 9 quant strategies with published methodology. 6 production AI projects. 31-gate eval harness with public kill log. Numbers below are counted at deploy time, not rounded. ArtifactTypeCountWhereQuant strategies (financial markets)alpha + risk, backtest, paper-trade · equities, futures, FX, crypto, alt-data9[/projects](/portfolio/projects/)Strategy workbooks + trade logsper-strategy memos, OOS metrics, kill log, drawdown windows9 + trade log[/workbooks](/portfolio/workbooks/)AI engineering projectsRAG · LLM evaluation · MCP · multi-agent systems for trading ops6[/ai](/portfolio/ai/)Public repos (GitHub)workbooks, ships, experiments25[github.com/christianmacion26](https://github.com/christianmacion26)Research validation methods (G1-G31)Walk-forward validation · block bootstrap · deflated Sharpe · regime conditioning31[/methodology](/portfolio/methodology/)Python codebaseBacktests · AI tooling · research infrastructure76.5k[/proof](/portfolio/proof/)Certs earned (last 11 months)AWS, GCP, quant, AI102[/certifications](/portfolio/certifications/) [ALL 96 PROJECTS →](/portfolio/projects/) 03 / certifications STRATEGY VALIDATION · 5-ERA STABILITY ## 5 eras. 17,700 trades. Deflated Sharpe 0.42 to 1.18 across regimes. Out-of-sample walk-forward validation. Block-bootstrap CIs. CSCV-based PBO. All 5 eras positive. BULL CRASH FRENZY HIKING NORMALIZE EraPeriodRegimeN tradesSharpeMax DDEra 12018-2019pre-COVID bull4,2001.42-8.1%Era 22020COVID crash + recovery2,8002.18-14.3%Era 32021meme / retail frenzy3,1000.94-11.2%Era 42022rates hiking3,5001.71-9.8%Era 52023-2024regime normalize4,1001.55-7.4%Aggregate2018-20245 regimes17,7001.56 avg-10.2% avg 00. gate lattice 31 INDIVIDUAL GATES · 6 FAMILIES · ONE CONTRACT G1 to G5Mechanical validityG1Schema conformanceG2Exit-0 contractG3Structured-output parseG4JSON Schema strict-modeG5Byte-length budgetG6 to G10Statistical nullsG6Block-bootstrap CIsG7Random-timing nullsG8Regime-shuffleG9Prior sensitivityG10Monte-Carlo coverageG11 to G15Walk-forward & OOSG11Locked OOS windowsG125-era stabilityG13Expanding walk-forwardG14Frozen-spec evaluationG15Embargoed test setG16 to G20Multiple testingG16Deflated Sharpe RatioG17CSCV PBOG18MinBTLG19Bonferroni-HolmG20BH-FDRG21 to G25Look-ahead disciplineG21Point-in-time datasetG22Survivorship-bias checkG23Rebalance-vs-signal timingG24Frozen-spec replayG25No-leakage to scoringG26 to G31Economic & cost realismG26Spread & slippageG27Capacity-constraint reportG28Funding-carryG29Borrow costG30Vol-targeted sizingG31Regime overlay § 03. Certifications ## 102 certifications · top 10 below. Tier-1 and flagship issuers only. Each row links to a verifiable issuer registry or credential page where available. The full list. every issuer, every date, every cert ID. lives on the[certifications page](/portfolio/certifications/). Top 10All 102 01 _(Meta logo)_ BIDA META AICCELERATE 2025 Meta · 2025Flagship [Verify ↗](https://www.bayanacademy.org/events/bida-year-2-meta-aiccelerate-training-2025-program) 02 _(NASA logo)_ Galactic Problem Solver : NASA Space Apps Challenge (Zurich, CH) NASA · 2025Flagship verify URL pending 03SO Certified Technical Analyst Program Society of Technical Analysts of the Philippines (Tier-1) · 2025Flagship [Verify ↗](https://staphilippines.org/) 04 _(Bitget logo)_ Blockchain4Youth (B4Y-2026-000701) Bitget · 2026Tier 1 [Verify ↗](https://www.bitget.com/promotion/blockchain4youth) 05D Introduction to AI Agents DataCamp · 2026Tier 1 verify URL pending 06F JPMorganChase : Investment Banking Job Simulation Forage · 2026Tier 1 verify URL pending 07 _(IBM logo)_ Artificial Intelligence Fundamentals IBM · 2026Tier 1 [Verify ↗](https://www.credly.com/org/ibm/badge/artificial-intelligence-fundamentals) 08UO Guest Speaker : CBA Annual Business Expo 2026 University of Southeastern Philippines (USeP) · 2026Tier 1 verify URL pending 09AD AI for the Modern Workforce: From Blockchain to the MOVE Language Ateneo de Davao University · 2025Tier 1 verify URL pending 10D Financial Trading in Python DataCamp · 2025Tier 1 verify URL pending [+ 92 more verified certifications →](/portfolio/certifications/) Full list → [/portfolio/certifications/](/portfolio/certifications/) IBM·Artificial Intelligence Fundamentals·2026 Cisco·AI Fundamentals with IBM SkillsBuild·2025 DataCamp·AI Fundamentals·2025 Google·Introduction to AI·2025 DataCamp·Introduction to AI Agents·2026 Ateneo de Davao University·AI for the Modern Workforce: From Blockchain to the MOVE Language·2025 Cisco Networking Academy·Introduction to Modern AI·2025 Cisco·Introduction to Data Science·2024 Cisco·Python Essentials 1·2024 DICT·Software Development and Design Thinking·2024 Cisco·Apply AI: Analyze Customer Reviews·2025 Bitget·Blockchain4Youth (B4Y-2026-000701)·2026 Society of Technical Analysts of the Philippines (Tier-1)·Certified Technical Analyst Program·2025 CertifyMe·Certified Technical Analyst·2025 Goldman Sachs·Completed Goldman Sachs Asset Management educational module: Foundations of Growth Equity (GS Alternatives University, 2025)·2025 Forage·JPMorganChase : Investment Banking Job Simulation·2026 DataCamp·Financial Trading in Python·2025 DataCamp·Math for Finance Professionals·2025 DataCamp·Intermediate Python for Finance·2025 DataCamp·SQL for Finance Professionals·2025 LinkedIn·Financial Modeling and Forecasting Financial Statements·2024 LinkedIn·Financial Forecasting with Analytics Essential Training·2024 LinkedIn·Using Data in Financial Analysis·2024 LinkedIn·Excel for Financial Planning and Analysis (FP&A)·2024 upGrad·Economics Masterclass·2024 Marginal Revolution University·Principles of Economics : Macroeconomics·2024 DataCamp / Khan Academy / university curriculum·Probability & Statistics Foundations·2024 DataCamp·Data Literacy·2025 DataCamp·Marketing Analytics for Business·2024 SEC Philippines·Fundamentals of Accounting·2024 SEC Philippines·Introduction to Capital Market·2024 SEC Philippines·Financial Reporting·2024 SEC Philippines·Introduction to Corporation·2024 Bangko Sentral ng Pilipinas·Financial Planning·2023 Bangko Sentral ng Pilipinas·Digital Financial Literacy·2023 Basel Institute on Governance·Operational Analysis of Suspicious Transaction Reports·2024 NASA·Galactic Problem Solver : NASA Space Apps Challenge (Zurich, CH)·2025 Meta·BIDA META AICCELERATE 2025·2025 University of Southeastern Philippines (USeP)·Guest Speaker : CBA Annual Business Expo 2026·2026 Civil Service Commission (Philippines)·Career Service Professional Eligibility·2023 HP LIFE·Effective Leadership·2024 HP LIFE·Agile Project Management·2024 University of the Philippines·Introduction to IT Project Management·2024 HP LIFE·Customer Experience (CX) for Business Success·2024 HP LIFE·AI for Business Professionals·2025 Alison·Diploma in Cryptocurrency·2024 Binance Academy·Crypto Trading Deep Dive·2024 AnChain.AI·Introduction to Cryptocurrency Recovery·2025 AnChain.AI·Fundamentals of Blockchain·2025 [ALL 102 CERTS →](/portfolio/certifications/) 04 / now ## § 04. What I Ship Four engagement shapes. Not services I list. Live systems built, deployed, kept honest with verifiable evidence. QUANT RESEARCH ### Statistical-arb pipelines that survive live deployment Walk-forward backtests, deflated Sharpe, regime conditioning. The 31-gate stack filters out edges that only looked good in-sample. Each pipeline ships with a paper-trade leg and a postmortem. AI ENGINEERING ### Multi-agent build systems with eval-first discipline Production agents behind a frozen eval harness and a kill-switch. Failures banked into postmortems; wins banked into the next eval set. Ship behind metrics, not vibes. AI EVAL ### Statistical gates for LLM claims Deflated Sharpe, block-bootstrap CIs, walk-forward stability. Same tools you use on backtests, applied to LLM eval scores. Built when the metric is too pretty. REPRODUCIBILITY ### Reproducible artifacts with audit trails Frozen-spec eval, idempotent re-run, one-page AAR. Verifiable by a stranger, not just an insider. [ALL PROOFS →](/portfolio/proof/) ## § 05. Footprint 20 venues across 3 regions. Real counterparts, real exchanges, real alt-data feeds. Verifiable, not decorative. APAC Digos City, PH6.74°N · 125.36°Ehome base · UTC+8 Manila, PH14.60°N · 120.98°EPDEx · PSE Singapore1.35°N · 103.82°ESGX · APAC derivatives Hong Kong22.32°N · 114.17°EHKEX · HSI · Hang Seng Tokyo35.68°N · 139.69°EJPX · TSE · Nikkei Shanghai31.23°N · 121.47°ESHFE · SSE Seoul37.56°N · 126.97°EKRX · Kospi Sydney33.87°N · 151.21°EASX Mumbai19.08°N · 72.88°ENSE · BSE · Nifty Dubai25.20°N · 55.27°EDGCX · regional oil EMEA London51.51°N · 0.13°WLSE · ICE · LCH · FX Frankfurt50.11°N · 8.68°EEurex · DAX Zurich47.38°N · 8.54°ESIX · CTA Zug, CH47.17°N · 8.51°ECrypto Valley · Ethereum AMERS New York40.71°N · 74.01°WNYSE · Nasdaq · Bloomberg · Refinitiv Chicago41.88°N · 87.63°WCME · CBOT · agricultural Toronto43.65°N · 79.38°WTSX · Toronto-Dominion San Francisco37.77°N · 122.42°WCoinbase · Kraken · FalconX São Paulo23.55°N · 46.63°WB3 · Bovespa GMU (GDELT)38.83°N · 77.31°WGDELT · alt-data · 15-min events --- # For recruiters · Christian T. Macion > The 60-second candidate one-pager: availability, location, target seats, resume PDFs, and direct contact for hiring managers. Source: `/for-recruiters/` FOR HIRING MANAGERS · 60-SECOND SKIM · NDA-CLEANChristian T. Macion · Digos City, Davao del Sur, Philippines (UTC+8) · UTC+8 · IMMEDIATE START CHRISTIAN.T.MACION·31-GATE EVAL·11-AGENT OFFICE·76.5k LOC PYTHON # 60-second skim. NDA-clean. Decide in one read. Tailored resume within 24 hours of a JD. Senior Quant Researcher and AI Engineer for financial markets. Manual-only posture. UTC+8. Remote-first. Looking for senior QR, QR-rotational, or research-engineer seats with US-premarket or APAC overlap. 6 paper-traded alpha strategies, 6 production AI projects, 25 public GitHub repos, and a 31-gate eval harness with a public kill log. Tailored resume within 24 hours of a JD. LOOKING FOR Senior QR or research-engineer seat with a real eval bar. Strategy team that ships, with public-data or self-owned research. AI engineer seat inside a quant team that builds LLM tooling. NOT LOOKING FOR Junior, intern, or rotational-only seats. Junior framing will end the conversation. Need comp ranges first? [See engagement bands →](/portfolio/comp/) StatusOpen · immediate start · 30 hrs/wkTarget seatsQuant Researcher (primary) · QR-Rotational · Quant Research Engineer · Research EngineerLocationDigos City, Davao del Sur, Philippines (UTC+8) · working UTC+8 · remote-firstWork authorizationFilipino citizen · independent contractor · globally remote (EOR) · no sponsorship requiredStackPython (76.5k LOC) · 11 agents · 31-gate eval harness · MCP · multi-agent LLM · RAG · frozen-spec evalsResume PDF[Tailored resumes →](/portfolio/resume/) ·[Quant (1-pg) ↓](/portfolio/resume-tailored.pdf) ·[AI-only ↓](/portfolio/resume-ai-only.pdf) ·[Unified ↓](/portfolio/resume-quant-onepage.pdf)Contact[christianmacion26@gmail.com](mailto:christianmacion26@gmail.com?subject=Inquiry%20%E2%80%94%20Christian%20Macion)· +63 991 616 2630 · [LinkedIn](https://www.linkedin.com/in/christianmacion)[Email Christian T. Macion →](mailto:christianmacion26@gmail.com?subject=Inquiry%20%E2%80%94%20Christian%20Macion)[View resumes →](/portfolio/resume/)[See proof →](/portfolio/proof/)[Cold-email template ↓](#cold-email-template) intro/ ## Cold-email template Paste, fill, send. Two mailto links pre-fill subject + body. Reply SLA: 4 business hours during UTC+8 day. Subject: [{role-tag}] {seniority} · JD attached to {company} Hi Christian : I’m a {your-role} at {company}, sourcing for a {role} seat. The JD is attached. Highlights: - {JD bullet 1} - {JD bullet 2} - {JD bullet 3} If you’re open, I’d like a 30-min screening this {Tue|Wed|Thu} between 18:00 to 22:00 PHT (UTC+8). Full agenda: christianmacion-portfolio.pages.dev/screening-call If not, who do you recommend? : {name} [Email [QR] track →](mailto:christianmacion26@gmail.com?subject=%5BQR%5D%20Senior%20%C2%B7%20JD%20attached%20to%20%7Brole%7D&body=Hi%20Christian%20%3A%0A%0AI'm%20a%20%7Byour-role%7D%20at%20%7Bcompany%7D%2C%20sourcing%20for%20a%20%7Brole%7D%20seat.%0AThe%20JD%20is%20attached.%20Highlights%3A%0A%20%20-%20%7BJD%20bullet%201%7D%0A%20%20-%20%7BJD%20bullet%202%7D%0A%20%20-%20%7BJD%20bullet%203%7D%0A%0AIf%20you're%20open%2C%20I'd%20like%20a%2030-min%20screening%20this%20%7BTue%7CWed%7CThu%7D%20between%2018%3A00-22%3A00%20PHT%20(UTC%2B8).%0AFull%20agenda%3A%20https%3A%2F%2Fchristianmacion-portfolio.pages.dev%2Fscreening-call%0A%0AIf%20not%2C%20who%20do%20you%20recommend%3F%0A%0A%3A%20%7Bname%7D) [Email [AI] track →](mailto:christianmacion26@gmail.com?subject=%5BAI%5D%20Eval-first%20%C2%B7%20JD%20attached%20to%20%7Brole%7D&body=Hi%20Christian%20%3A%0A%0AI'm%20a%20%7Byour-role%7D%20at%20%7Bcompany%7D%2C%20sourcing%20for%20a%20%7Brole%7D%20seat.%0AThe%20JD%20is%20attached.%20Highlights%3A%0A%20%20-%20%7BJD%20bullet%201%7D%0A%20%20-%20%7BJD%20bullet%202%7D%0A%20%20-%20%7BJD%20bullet%203%7D%0A%0AIf%20you're%20open%2C%20I'd%20like%20a%2030-min%20screening%20this%20%7BTue%7CWed%7CThu%7D%20between%2018%3A00-22%3A00%20PHT%20(UTC%2B8).%0AFull%20agenda%3A%20https%3A%2F%2Fchristianmacion-portfolio.pages.dev%2Fscreening-call%0A%0AIf%20not%2C%20who%20do%20you%20recommend%3F%0A%0A%3A%20%7Bname%7D) Reply SLA · 4 business hours during UTC+8 day. Cold DMs: 24h SLA. ## FAQ 4 canonical questions. Short, factual, verifiable from the rest of the site. 01What seniority level are they targeting? Senior / staff Quant Researcher or Research Engineer. Primary seat: Quant Researcher (with QR-rotational considered). Secondary: AI Engineer inside a quant team that ships LLM tooling. 02What data do the published results use? Public sources only: Yahoo Finance, CoinGecko, FRED, EDGAR, SEC filings, public CME/CBOE chains. Every result on this site reproduces from them. 03How fast can they respond? Four business hours during the UTC+8 day for recruiter email. Cold DMs slower. Direct phone for qualified screening calls only. 04Do they require sponsorship? Not for remote engagements. the candidate works as an independent contractor from the Philippines. For full-time W-2 roles, visa sponsorship is the employer-side question; the candidate holds a Philippine passport and is open to globally remote (EOR) arrangements. --- # Proof · Christian T. Macion > Results with the artifact behind each one: 9 quant projects and 6 AI projects, metric per row, public-data reproducible. Source: `/proof/` PROOF · SHOW YOUR WORK15 ARTIFACTS · PUBLIC-DATA REPRODUCIBLE CHRISTIAN.T.MACION·UTC+8·16 ARTIFACTS·NDA-CLEAN·OWNER-VERIFIED # Proof. Public or self-owned only. 15 artifacts. Public-data reproducible. 16 on disk. Strongest evidence first: 9 quant strategies with public-data backtests, then6 AI projects with frozen-spec eval harnesses. Each row links to the write-up. Metrics come from project frontmatter, not hand-written. On this proof page [Research artifacts](#research) [Track record](#track) [Public track record](#public-track) [Evidence index](#evidence) [NDA disclosure](#disclosure) [Live terminal→ /markets](/portfolio/markets/) P.2 · Research artifacts ## 9 quant · 6 AI. Public-data reproducible. One row, one result. Metric from project frontmatter. Each row links to the write-up. TopicAllQuantAIOSSTypeAllCodeWriteupTalkLive demo ### Quant projects to 9 Open Source·2025-09-20·12 min read[QUANTMultiple Testing & the Deflated Sharpe Ratio1.14Best-of-160 in-sample SharpeSep 2025→](/portfolio/projects/quant/01-deflated-sharpe/) Best-of-160 BTC rule: IS Sharpe 1.14 is only 1.24× the pure-noise expectation : DSR = 0.70 (fail). Open Source·2025-10-15·10 min read[QUANTCross-Sectional Momentum (18 coins)0.91In-sample SharpeOct 2025→](/portfolio/projects/quant/02-cross-sectional-momentum/) IS Sharpe 0.91 → OOS −0.03 : an honest decay; bootstrap CI straddles zero. Open Source·2025-12-20·9 min read[QUANTTime-Series Momentum + Vol Targeting0.27Sharpe (raw)Dec 2025→](/portfolio/projects/quant/03-timeseries-momentum-voltarget/) Vol-targeting lifts Sharpe 0.27 → 0.39 and halves max DD (−62% → −30%). Open Source·2026-01-25·11 min read[QUANTThe Variance Risk Premium (VIX vs Realized)85%Years VRP > 0Jan 2026→](/portfolio/projects/quant/04-variance-risk-premium/) Implied > realized 85% of 36 yrs; predicts returns, Newey-West t = +6.5. Open Source·2026-02-18·8 min read[QUANTPairs Trading via Cointegration (BTC / ETH)−2.11ADF test statisticFeb 2026→](/portfolio/projects/quant/05-pairs-cointegration/) ADF −2.11 (not cointegrated), half-life 208 d : the fade loses, as the test predicts. Open Source·2026-04-05·10 min read[QUANTCrypto Funding-Carry+11.9%Annualized fundingApr 2026→](/portfolio/projects/quant/06-funding-carry/) Funding +11.9% annualized premium; fade IS 1.11 → OOS −0.05 (decayed post-2023). Open Source·2026-05-10·9 min read[QUANTMacro / Volatility-Regime Overlay0.64Sharpe (base)May 2026→](/portfolio/projects/quant/07-macro-regime-overlay/) Vol-managed exposure: Sharpe 0.64 → 0.67, max DD −58% → −42%. Open Source·2026-05-28·11 min read[QUANTBacktest Engine + Cost Model0.20Sharpe (gross)May 2026→](/portfolio/projects/quant/08-backtest-engine-costs/) High-turnover signal wins gross (0.20 > 0.13) but loses net of cost : break-even 20 bps. Open Source·2026-06-18·8 min read[QUANTLook-Ahead Bias Audit (the shift test)0.59Clean SharpeJun 2026→](/portfolio/projects/quant/09-lookahead-bias-audit/) A leak inflates Sharpe 0.59 → 5.07; the shift test exposes it as 88% phantom. ### AI projects to 6 Open Source·2025-11-15·8 min read[AIRAG Recall Eval0.886recall@3Nov 2025→](/portfolio/projects/ai/01-rag-recall/) RAG service that proves its own retrieval : recall@3 = 0.886, MRR@3 = 0.805 with offline stdlib TF-IDF retriever. Open Source·2026-02-04·9 min read[AIEval MCP Server20 / 20MCP conformanceFeb 2026→](/portfolio/projects/ai/04-eval-mcp-server/) MCP server exposing the slop-evaluation gate over Tools, Resources, and Prompts : 20/20 conformance, 100% round-trip parity. Open Source·2025-12-08·10 min read[AITool-Call Agent100%Tool / arg correctnessDec 2025→](/portfolio/projects/ai/02-toolcall-agent/) ReAct-style tool-calling agent with OTel traces, fault injection, and 100% tool/arg correctness. Open Source·2026-01-12·11 min read[AILLM-as-Judge Harness0.58Cohen's κ (vs human)Jan 2026→](/portfolio/projects/ai/03-judge-harness/) LLM-as-judge pipeline validated against human raters : Cohen's κ = 0.58 with bootstrap CI and position-bias measured. Open Source·2026-03-01·7 min read[AIReflect-Revise Loop127.5 → 14.0Mean SLOP scoreMar 2026→](/portfolio/projects/ai/05-reflect-revise/) Reflection-loop agent : mean SLOP 127.5 → 14.0 across drafts, 3/4 improved, 1/4 honest no-progress halt. Open Source·2026-03-22·6 min read[AISlop Scanner81 → 3Real-draft improvementMar 2026→](/portfolio/projects/ai/06-slop-scanner/) 13-metric literature-grounded AI-output quality gate : drove a real draft from HEAVY (81) to CLEAN (3). P.5 · Public track record ## Lectures, charts, feeds. Two USeP EGE 313 lectures, one public TradingView chart, three Atom feeds (portfolio · project · solution updates). Live Presenter-view delivery, Sep 13 2024. Webcam lower-right; deck filename 'MACION. MANOBO. PIC. ECE3A. 230pm. MW' visible in the PPT Edit-mode frame.Three weeks before the lecture: same deck, drafted end-to-end via a ChatGPT session. The precursor pattern to the current 31-gate authoring harness. ### Operator-discretion chart _(TradingView chart of XAUT/USDT (tokenized gold) on the 1-hour timeframe with Elliott Wave counts. Public chart by @CryptoneedsChrist.)_ XAUT/USDT perpetual futures, 1h, MEXC. Elliott Wave count by @CryptoneedsChrist (public TradingView handle). Operator-discretion; not part of the systematic book.@CryptoneedsChrist · public TradingView · Jan 3 2026 ### Atom feeds. 3 streams [/portfolio/feed.xml](/portfolio/feed.xml) · portfolio updates [/portfolio/feed-projects.xml](/portfolio/feed-projects.xml) · project updates [/portfolio/feed-solutions.xml](/portfolio/feed-solutions.xml) · solution updates P.5b · External accounts Six external accounts and surfaces, each with a handle and a public, dated metric. Hairline-rule rows, no thumbnails. If a row is empty, it means that channel isn't public yet. Numerainot public yetcorrelation / MMC / TC[see /positions](/portfolio/positions/)HuggingFace@christianmacionMCP server artifact[/projects/ai/04-eval-mcp-server](/portfolio/projects/ai/04-eval-mcp-server/)GitHubchristianmacion26102 certs · 76.5k LOC · 15 OSS projects[github.com/christianmacion26](https://github.com/christianmacion)TradingView@CryptoneedsChristpublic chart · Elliott Wave[tradingview.com/u/CryptoneedsChrist](https://www.tradingview.com/u/CryptoneedsChrist/)Atom feeds3 streamsfeed.xml · feed-projects.xml · feed-solutions.xml[/feed.xml](/portfolio/feed.xml)Medium@christianmacion26public essays[medium.com/@christianmacion](https://medium.com/@christianmacion) P.5d · Evidence index ## The AARs behind the page. Twelve AARs from the last 30 days that produced, audited, or updated the artifacts above. Each filename resolves to a markdown file at ~/.claude/cache/corporate/aars/ . DateAARWhat it produced2026-08-01portfolio-v9-3-wave-2-shipWave-2 ship: chrome, motion, a11y, IA. the v9.3 baseline this page sits on2026-08-01portfolio-v9-3-wave-2.5-security-fixesWave-2.5: HSTS + 5 CVE patches + Lighthouse 4/4 97-100, audit 64→88+2026-08-01portfolio-v9-3-architecture-auditArchitecture audit · score 62 · the basis for C-1 ChromeHeader consolidation2026-08-01portfolio-v9-3-voice-auditVoice audit · score 88 · the basis for P.5c 'pending attestation' rewrite2026-08-01portfolio-v9-3-a11y-auditA11y audit · score 95 · WCAG 2.2 AA + axe 0/912026-08-01portfolio-v9-3-token-dedupe-atomicToken dedupe · 52→0 drift · the basis for the sharp-corner register2026-08-01portfolio-v9-3-wave-5-verificationWave-5 verify · 9/9 routes 3/3 mobile · 1173/1173 QA baseline2026-08-02portfolio-design-auditThe v9.3.1 design audit this page rebuild closes (50 findings)2026-08-02accion-thesis-b-triageQuant strategy triage · Lane C gates applied (G40 to G45 + K13)2026-08-02c2-deep-preflightC2 deep pre-flight · READY-FOR-BACKTEST · 6 hard-stop kill triggers2026-08-02afk-sweep-v2p6-outcomeAFK sweep v2.6 · 10/10 KILL · 43 cumulative · real-but-too-thin binding constraint2026-08-02methodology-amendments-approvedG36 + G37 + R-methodology-loop approved · data-coverage + bid-ask-bounce P.6 · What's not here ## NDA-positive disclosure. What's not here: photos of NDA-protected offices. Proprietary strategy outputs. Client calendars. Everything on this page is public-data or self-owned. The systematic-strategy desk role is a closed past contract. If the proof you need isn't on this page, email me. If it's public, I send it within 24 hours. ## Want a specific artifact? Email me with the proof you need. If it's public, NDA-clean, and I have it on disk, I'll send it within 24 hours. [Email me](/portfolio/resume/)[Read /experience](/portfolio/experience/) --- # Methodology · Christian T. Macion > The eval-first discipline. G1 to G31 statistical gate stack. Model-routing policy. Agent charter pattern. Source: `/methodology/` METHODOLOGY · EVAL-FIRST DISCIPLINEG1 to G31 GATE STACK · MODEL-ROUTING · AGENT CHARTER CHRISTIAN.T.MACION·UTC+8·31-GATE EVAL·OWNER-VERIFIED # Don’t ship what hasn’t passed a measurable gate. Five stages. Each kills 70%+ of the work. Same gate stack rates systematic strategies and LLM outputs. Regime. Classify market state. Trend / range / vol regime before any signal. Pre-flight. Mechanical validity. Schema, parse, dedupe, exit-0. Research. Walk-forward, deflated Sharpe, block-bootstrap, orthogonality. Backtest. OOS, regime conditioning, Monte-Carlo ruin probability. Scribe. One-page memo + kill log. Verdict derives from gates. Every shipped primitive passes through 31 gates: mechanical validity, statistical nulls, walk-forward OOS, multiple testing, look-ahead discipline, economic and cost realism. Same gate IDs mean the same thing whether the artifact is a trading strategy or an LLM output. Pr(ship)=Pr(ship∣passes G1–G31)⋅i=1∏31​Pr(Gi​∣evidencei​)On this methodology page [G1 to G31 gate stack](#gates) [Model routing](#routing) [DSR calculator](#dsr) [Agent charter](#charter) 00 / methodology The gate stack ## G1 to G31. Six families. One contract. Every candidate (LLM output, systematic strategy, content artifact) passes through the same gate stack. Gates are mechanical (exit-0 contract), not opinionated. A pass means the candidate is allowed to surface; a fail means it isn't. Gate IDs are load-bearing: a "G16" means the same thing across the corpus. Gate IDFamilyWhat it checksG1 to G5Mechanical validity Schema conformance, exit-0 contract, structured output parse, JSON Schema strict-mode, byte-length budget. G6 to G10Statistical nulls [block-bootstrap](/portfolio/glossary/block-bootstrap/) CIs, random-timing nulls, [regime](/portfolio/glossary/regime/)-shuffle, parameter-prior sensitivity, Monte-Carlo coverage. G11 to G15Walk-forward & OOS Locked [OOS](/portfolio/glossary/oos/) windows, 5-era stability, expanding/rolling [walk-forward](/portfolio/glossary/walk-forward/) windows, frozen-spec evaluation, [embargoed](/portfolio/glossary/embargo/) test set. G16 to G20Multiple testing [Deflated Sharpe Ratio](/portfolio/glossary/deflated-sharpe/) (DSR), [CSCV-based Probability of Backtest Overfit](/portfolio/glossary/cscv-pbo/) (PBO), [Minimum Backtest Length](/portfolio/glossary/minbtl/) (MinBTL), [Bonferroni-Holm](/portfolio/glossary/bonferroni-holm/), BH-FDR. G21 to G25Look-ahead discipline Point-in-time dataset verification, [survivorship-bias](/portfolio/glossary/survivorship-bias/) check, rebalance-vs-signal timing, frozen-spec, no-leakage to scoring. G26 to G31Economic & cost realism Spread/[slippage](/portfolio/glossary/slippage/)/latency per asset class, capacity-constraint report, funding-carry, borrow cost, vol-targeted sizing, [regime](/portfolio/glossary/regime/) overlay. 00. gate lattice 31 INDIVIDUAL GATES · 6 FAMILIES · ONE CONTRACT G1 to G5Mechanical validityG1Schema conformanceG2Exit-0 contractG3Structured-output parseG4JSON Schema strict-modeG5Byte-length budgetG6 to G10Statistical nullsG6Block-bootstrap CIsG7Random-timing nullsG8Regime-shuffleG9Prior sensitivityG10Monte-Carlo coverageG11 to G15Walk-forward & OOSG11Locked OOS windowsG125-era stabilityG13Expanding walk-forwardG14Frozen-spec evaluationG15Embargoed test setG16 to G20Multiple testingG16Deflated Sharpe RatioG17CSCV PBOG18MinBTLG19Bonferroni-HolmG20BH-FDRG21 to G25Look-ahead disciplineG21Point-in-time datasetG22Survivorship-bias checkG23Rebalance-vs-signal timingG24Frozen-spec replayG25No-leakage to scoringG26 to G31Economic & cost realismG26Spread & slippageG27Capacity-constraint reportG28Funding-carryG29Borrow costG30Vol-targeted sizingG31Regime overlay 01 / trend arc Why the gate stack tracks the curve ## The methodology is not a static document. Each rung of the AI arc changed how the gate stack is wired. PROMPTS gave the eval family; SKILLS gave the structured-output contract; AGENTS gave the run-loop schema; HARNESS gave the sandbox; LOOPS gave the AAR pattern; GRAPHS gave the shared-memory graph that connects every gate test to every shipped artefact. OFFICES is where the gate stack now runs: 14 squads, 112 agents, every gate test attached to a graph node. Below: the dated record of adoption versus mainstream emergence. AI · HISTORY · METHODOLOGY EVIDENCE ## Six rungs that built this gate stack. The methodology is wired rung-by-rung. Each architectural jump maps to a gate family on this page. 11MONTHS AHEAD·7STAGES·2022 to 2025ARC·112AGENTS NOW PROMPTS Nov 2022→Dec 2022 I adopted Dec 2022 1months ahead Single-turn Q&A; research summarization SKILLS Jul 2023→Aug 2023 I adopted Aug 2023 1months ahead Tools/functions; structured output AGENTS Dec 2023→Feb 2024 I adopted Feb 2024 2months ahead Single-agent loops; tasks HARNESS Apr 2024→Jun 2024 I adopted Jun 2024 2months ahead AI writes code in loop LOOPS Sep 2024→Nov 2024 I adopted Nov 2024 2months ahead Persistent agentic workflows GRAPHS Mar 2025→May 2025 I adopted May 2025 2months ahead Persistent state + memory; knowledge graphs OFFICES Q4 2025→Q4 2025 I adopted Q4 2025 1months ahead 14 squads · graph-engineered multi-agent orchestration NOW [Full timeline →](/portfolio/timeline/) Model routing ## Tiered dispatch, not single-agent. A multi-agent system runs at ~15× the token cost of a single agent. Route by what the step needs to compute, not by default. Every dispatch has a per-step budget. Opusjudgment / hardest autonomous decisions~1 dispatch per analysis chainSonnetassembly / structured generation~3 to 5 dispatches per chainHaikumechanical / fast deterministic steps~10 to 30 dispatches per chainFablelong-horizon autonomous researchone off, overnightcost=∑i​tokens(i)⋅price(tier(i))wheretier:A→{Opus, Sonnet, Haiku, Fable} end-to-end pipeline · every project 01Intentproblem statement + success…02Framefrozen spec · eval harness…03Buildtier-routed agents · cost-c…04Verifygate stack G1 to G31 · exit…05Deliverpublic repo or live paper-t…↻ this loops. feedback from deliverable revises intent01Intentproblem statement + success…02Framefrozen spec · eval harness…03Buildtier-routed agents · cost-c…04Verifygate stack G1 to G31 · exit…05Deliverpublic repo or live paper-t…↻ feedback loop: deliverable revises intent Interactive · G16 ## Try the [deflated Sharpe](/portfolio/glossary/deflated-sharpe/) calculator. Same in-sample [Sharpe](/portfolio/glossary/sharpe/), different DSR, depending on how many strategies you tried. This is the gate that turns "1.14 looks great" into["DSR 0.70, fail."](/portfolio/glossary/deflated-sharpe/) deflated Sharpe · interactive ### DSR calculator Drag the sliders. DSR tells you whether your in-sample Sharpe survives the multiple-testing penalty for the number of strategies you tried. Strategies tried 16011001000In-sample Sharpe 1.14−0.52.04.0Years of data 4.00.5y10y20ycomputing…n/aDSRdenominator n/a expected SR n/a γ₃ n/a γ₄ n/a 01.5 (pass)3.0 DSR uses the normal-approximation of the expected maximum Sharpe under N i.i.d. trials, then deflates by √(1 − γ₃·SR + (γ₄·SR² − 1)/4). Why 0.95? to reading the z-score and the numerator The number the calculator prints is a z-score on the null H₀: "your IS Sharpe is just the best of N i.i.d. N(0,1) draws." Pass at DSR ≥ 1.645 = one-sided 95% test (the standard frequentist threshold; equivalent to 0.05 significance). Anything below is inconclusive or fail. not because 0.05 is sacred, but because 1.645σ is the conventional bar for "I tried many things and this one still looks unlikely to be noise." Read it as a ratio: numerator = SR̂ − E[max SR under N trials] (how much your Sharpe beats the expected best of N random strategies); denominator = σ_SR̂ · √(1 − γ₃·SR + (γ₄−1)·SR²/4) (how noisy the Sharpe estimate is, with the higher-moment correction for non-normal returns). A DSR of 0.70 means "the gap is 0.70 noise-units wide" to not a 70% pass rate. Ref: Bailey & López de Prado, "The Deflated Sharpe Ratio: Correcting for Selection Bias, Backtest Overfitting, and Non-Normality" (2014), JPM 40(5). Agent charter ## 6 required fields per agent. A charter is a contract, not a prompt. Same six fields, same semantics, every agent. No exceptions. roleA specific function, never generic "AI assistant"inputsTyped input contract (JSON Schema, fail-closed)outputsTyped output contract with strict validationhandoffsWhat this agent must NOT do (anti-charter)gatesWhich of G1 to G31 apply, with exit codesevidenceWhat the agent must attach as proof Related pages ## Where the rest of the method lives. This page owns the gate stack, model routing, and the agent charter. The incident log (where the gate stack caught a real bug) lives on[/mistakes](/portfolio/mistakes/). The first-90-days plan for a new QR seat is on[/for-recruiters](/portfolio/for-recruiters/). The implementation record (repos, licenses, verification dates) lives on [/colophon](/portfolio/colophon/). --- # Now · Christian T. Macion > What Christian is shipping this month (9 August 2026). A dated work log, updated on ship. Source: `/now/` 01 — NOW · LAST UPDATED 2026-08-09 # What's shipping this month Dated. Honest. Manual-only posture. Updated only when something changes that matters. Live trading terminal running manual-only on a $2K Upcomer Vanguard account. Stage 2.5 multi-asset sweep: 37 cells scored at 15m resolution, 0 survived FDR at q=0.10. Vanguard 2K synthesis: 0/26 cleared. Walk-forward OOS standardization rolling out across 8 quant projects by end of August. ## Currently shipping Bloomberg-style trading terminal. 9-panel layout for the live Upcomer Vanguard $2k account. Manual-only posture binding. Yahoo + Binance live data. 6-hour AFK build, CONVERGED 2026-08-08. Stage 2.5 multi-asset sweep. 37 cells scored at 15m resolution across XAU/XAG/BTC/ETH. 11 carry the G16 monotone-positive fingerprint, 0 survive FDR at q=0.10. Resolution-invariant. Vector coverage is the constraint. Vanguard 2K synthesis. 0/26 cleared (BTC 3/3 + Gold 19/19 + Silver 3/3 KILL). Signal firer dormant, manual-only. Top-3 priorities queued. Portfolio v12.W4 Bryllim parity. 10-hub nav (6 icon+label + 4 plain text). 5 v12.W3 regression fixes. Headshot re-crop to home hero. Site-wide axe sweep queued. Walk-forward OOS standardization. Applying the locked-window pattern from project 06 to the remaining 8 quant projects. ETA end of August 2026. ## Currently reading Shreve, Stochastic Calculus for Finance II. Ch. 4. Itô lemma + change-of-numéraire. Anchors the vol-target time-series momentum framework. Bailey & López de Prado, Deflated Sharpe Ratio. Re-read for the G16 rule wording across the methodology page. Deming, Out of the Crisis. System of Profound Knowledge. Common cause vs special cause. Maps cleanly onto the 31-gate eval harness. ## Currently thinking about The vector-coverage gap. Stage 2 keeps killing on coverage, not edge. Worth adding a G-candidate for regime-stratified sample-size minimums before Stage 3 burns another sweep. Eval bars for LLM-eval claims. A single Brier + FDR pair beats every "vibe" leaderboard. Standardizing the wording so multi-agent eval pages cite the same gate IDs as quant research. Hiring context: [/for-recruiters](/portfolio/for-recruiters/) · reach me: [/contact](/portfolio/contact/) · commit log: [github.com/christianmacion26/portfolio](https://github.com/christianmacion26/portfolio/commits/main). --- # Live paper-trading · Christian T. Macion > Currently-running strategies. Public data, no live capital, weekly updates. Source: `/positions/` Status: open to conversations · last refreshed 2026-08-09 POSITIONS · LIVE PAPER-TRADINGNO LIVE CAPITAL · PUBLIC DATA · WEEKLY # What I'm currently running. Paper-trade first. Real capital only after the gates pass. never before. A backtest is what a strategy would have done . A paper-trade is what it actually does . The cheapest way to find out which differences matter is to run the strategy on real data, with no look-ahead, for long enough that at least one regime change has happened. This page is the canonical record of that run. No live capital · paper-trade only Public data only · Binance BTC/USDT 1d klines Weekly updates · or on regime-event override CHRISTIAN.T.MACION·UTC+8·LAST REFRESHED 2026-08-09·OPEN TO CONVERSATIONS·OWNER-VERIFIED 00 / positions POSITION 01 · ACTIVE ## Vol-target time-series momentum · BTC/USDT project 03 · running since 2026-04-21 · 50 trading days · vol target 30% annualized live · in regime Equity curve · drawdown · vol-regime Live paper-trade series, normalized to 100. Top panel: equity. Bottom panel: running drawdown (red fill). Background tint: rolling 5-day vol quantiles. calm / normal / stress. Net of 5 bps/side slippage. 12% return over 50 trading days ≈ 0.39 Sharpe annualized, matching the IS backtest. Vol-target sizing held max DD under 4% to lower than the 30% headline because the live period has been a calm regime. live series99.0102.5106.0109.5113.0-1.5%-1.0%-0.5%0.0%EQDDt=0t=25dt=49d calm (low vol) normal stress (high vol) regime from rolling 5d vol of the series itself Total return+12.0%Net of 5 bps/side costAnnualized Sharpe≈ 0.39Matches IS backtestMax drawdown−3.6%Lower than 30% headline (calm regime)Avg daily turnover≈ 0.4×Weekly rebalance cadenceRealized slippage≈ 4.2 bpsWithin cost-model bandCapacity est.$3 to 5MNotional at <5 bps impact ### Tradeoffs observed vs backtest Slippage matches: 4.2 bps/side realized vs 5 bps modeled. The linear-impact assumption is honest at this size. Regime stability: the period was dominated by a calm regime. The backtest's worst-drawdown number (−30%) is untested in this window. The next regime event will be the test. Latency: Binance kline timestamp drift is real but small at the 1-day cadence. Not a concern at this strategy frequency. Survivorship: only one instrument. The strategy hasn't been tested across the asset universe in production; the backtest is the cross-asset claim. ### What I'd change before going live Add a kill-switch on >20% weekly drawdown (current code has none. this is a backtest-engineering gap, not a strategy gap). Add a per-trade log with size, slippage realized, and execution timestamp for capacity testing at $1M / $5M / $10M tiers. Cross-check the signal against Binance.US and Coinbase klines to rule out venue-specific artifacts. code:[github.com/christianmacion/timeseries-momentum-voltarget](https://github.com/christianmacion/timeseries-momentum-voltarget) Open conversations · the live book ## Conversations I'd pick up the phone for. 5 seats that fit my operating constraints. quant, AI, eval, OSS. Not a JD wishlist; a list of the conversations I'd actually take a 30-min call about. If your role isn't on the list, email me anyway with the JD. I read every one. I do 30-min intro calls on Tue/Wed/Thu 18:00 to 22:00 PHT (10:00 to 14:00 UTC). Email first : booking link in reply. [See the 30-min agenda →](/portfolio/screening-call/) [Email for intro call →](mailto:christianmacion26@gmail.com?subject=Intro%20call%20request) CONVERSATION 01 ### Quantitative Researcher Hedge fund or family office · remote-first · US-premarket or APAC overlap Looking for Multi-strategy pod, eval-first research, async-friendly cadence. I'm a fit because 11-agent platform + 31-gate statistical eval harness, NDA-clean by construction. [Next step: /for-recruiters →](/portfolio/for-recruiters/) CONVERSATION 02 ### AI Research Engineer Frontier AI lab (HF / Replicate / HuggingFace / Conjecture) · remote · model eval + RAG infrastructure Looking for Engineer who ships eval harnesses + RAG scorecards, not just prompts. I'm a fit because Eval-first MCP, RAG scorecard, slop gate, 76.5k LOC Python. [Next step: /for-recruiters →](/portfolio/for-recruiters/) CONVERSATION 03 ### Quant Research Engineer Crypto prop desk / market-making shop · remote · public-data discipline Looking for Cost-model treated as a first-class artifact, not an afterthought. I'm a fit because Funding-carry, cost-aware backtests, public-data rigor (Binance / Coinbase / EDGAR / FRED). [Next step: /for-recruiters →](/portfolio/for-recruiters/) CONVERSATION 04 ### NLP Evaluation Engineer AI safety / evals org · remote · spec-driven methodology Looking for LLM-as-judge harness, position-bias checks, frozen-spec eval. I'm a fit because 31-gate eval harness, position-bias controls, frozen-spec discipline. [Next step: /for-recruiters →](/portfolio/for-recruiters/) CONVERSATION 05 ### Open-source Maintainer OSS foundation / grant-funded org · async · public-impact work Looking for Public, evaluable, reproducible artifacts. not private magic. I'm a fit because 25 public repos, 102 certs, eval-first discipline, NDA-clean by construction. [Next step: /for-recruiters →](/portfolio/for-recruiters/) Off the table ## Not currently looking for. 3 seat types that don't fit my operating constraints. No judgment if you run one. just not where I'm spending the year ahead. HFT colocation-dependent seats (Digos is 8 hours ahead of NYSE; latency kills the edge) Sell-side black-box vendor roles (I prefer open-data / open-source provenance) Crypto market-making seats that require < 100ms latency (same reason) What's not on this page ## Negative space as design. A single live paper-trade is honest. A dozen concurrent paper-trades without per-strategy slippage / capacity logs is vanity. This page is short on purpose. one position, fully documented, beats six positions hand-waved. I am not running live capital. Paper-trading only. I am not running on proprietary data. Binance BTC/USDT 1d klines are public. I am not claiming the live period validates the backtest. The next regime event will be the test. I am not running more strategies than I can document. One position, fully logged. ## Want the trade log or capacity report? Send the strategy you're hiring for. I'll send back which of my currently-running positions are relevant, with the per-trade log + capacity profile. [Email me](/portfolio/resume/)[Read the backtest](/portfolio/projects/quant/03-timeseries-momentum-voltarget/) --- # Experience · Christian T. Macion > Operator and researcher experience across multi-agent AI platforms, venture incubation, content automation, and systematic trading research. Source: `/experience/` 01. experience EXPERIENCE · ENGAGEMENTS + PROOFS8 ENGAGEMENTS · NDA-SAFE · 22-MONTH ARC CHRISTIAN.T.MACION·UTC+8·NDA-CLEAN·PROOF-PER-ROW # Where the work happened. and how to verify it. Each engagement ships with proof. If the proof isn't public, the entry isn't here. 8 engagements. Each entry lists the contributions made and points to public artifacts that prove them. The systematic-strategy desk role is a closed past contract (03/2026 to 06/2026); all references are NDA-safe and free of proprietary data sources. NDA-safe by construction. Every entry is built from public repos, public talks, and public-data backtests. No proprietary data sources, strategy internals, or desk PnL appear anywhere on this site. Missing proof? email me and I'll send it within 24h if it's public. contribution_density=months(ΔE)∑e∈E​proofs(e)​engagements∣E∣8multi-hatcontributions∑ce​32↑declaredproof pointers∑pe​22↑public artifactsmonthsΔt222024-09 → 2026-08 4 min read · last updated2026-08-09 · [how verified ↗](/portfolio/proof/) ## AI Systems Engineer / Quantitative Researcher (contract) systematic-strategy desk (NDA-protected; closed past contract 03/2026 - 06/2026)·Remote Mar 2026 to Jun 2026 Architected the 11-agent AI research platform and the 31-gate statistical eval harness used by the desk under a PM with publicly attributable initials. (Past contract; closed 06/2026.) Designed and operated an 11-agent orchestrator-worker AI research platform (~27,500 words of role-scoped agent charters) with contracted hand-off packets, few-shot routing, and separation-of-duties between generation, validation, and documentation agents : backed by ~76,500 LOC of light-dependency Python. Built a 31-gate statistical evaluation harness (G1-G31): block-bootstrap CIs, random-timing nulls, walk-forward, 5-era stability, plus a multiple-testing layer (Deflated Sharpe, CSCV-based PBO, Minimum Backtest Length) : implemented scipy -free in numpy and used to validate both LLM outputs and systematic-trading strategies. Architected a tiered model-routing policy (Opus = judgment / Sonnet = assembly / Haiku = mechanical / Fable = hardest autonomous) with per-dispatch token budgeting : the cost layer that keeps a multi-agent system (measured at ~15× single-agent token cost) economical to run continuously : plus structured-output contracts gated by a mechanical validator (must exit 0). Researched and validated systematic trading strategies across 5 asset classes (equity-index, crypto, energy, metals, agriculture) on the desk’s statistical-filter platform; shipped candidates through paper-shadow and live forward out-of-sample testing under pre-registration and frozen-spec evaluation. Built point-in-time, look-ahead-disciplined data pipelines from scratch on free public sources; implemented byte-range subsetting, completeness guards, gap logging, and idempotent incremental pulls to produce reproducible research-grade datasets. Caught and documented false positives as enforced methodology : banking each into the desk’s reusable research-integrity playbook : and designed an automated monthly forward-out-of-sample monitoring fleet (scheduled data pull → S3 sync → frozen-spec evaluation → ledger) that produces un-gameable live performance evidence. CONTRIBUTIONS · EVIDENCE · PROOF 11-agent orchestrator + 31-gate eval harnessPublic scorecards for the eval gates (G1-G31); methodology in publications log See /solutions/11-agent-eval-platform + methodology page Deflated Sharpe, CSCV-based PBO, MinBTLThree canonical multiple-testing corrections in one gate stack OSS repo (deflated-sharpe module); publications log entry 5 asset-class systematic researchLocked OOS windows, frozen-spec evaluation per candidate Memo + figure for each (public-data) Public-data PIT data pipelinebyte-range subsetting, gap guards, idempotent incremental pulls Pipeline code in public quant portfolio PROOF POINTERS ↗Contract disclosed under NDA : role: AI Systems Engineer / Quantitative Researcher. PM with publicly attributable initials (NDA-protected desk). ↗Eval harness methodology (G1-G31, deflated Sharpe, block-bootstrap CIs) carried forward into the public methodology page ↗All deliverables NDA-safe; no proprietary data sources referenced ## Founder & AI Systems Engineer Macion Ventures·Remote Jan 2025 to May 2026 Built a 7-agent venture-incubation pipeline (5 judgment-tier + 2 mechanical) with 10 lifecycle skills and an anti-self-approval governance pattern. Built a 7-agent venture-incubation pipeline (5 judgment-tier + 2 mechanical) with 10 lifecycle skills; produced 31 decision-grade artifacts and engineered an anti-self-approval governance pattern (the agent that proposes never approves). Encoded Philippine tax/regulatory rules (DTI/SEC/BIR/LGU, ₱3M VAT threshold, 8%-flat vs graduated election) directly into agent and skill prompts, so the research output is jurisdiction-aware at the prompt level : not bolted on at the end. Shipped 31 decision-grade artifacts (charters, briefs, ledger entries) under the same separation-of-duties principle used in production research desks. CONTRIBUTIONS · EVIDENCE · PROOF 7-agent venture-incubation pipeline5 judgment-tier + 2 mechanical agents, 10 lifecycle skills, 31 decision-grade artifacts See /solutions/7-agent-venture-pipeline Anti-self-approval governance patternThe proposing agent never approves its own output : separation of duties at the agent level Pattern documented in solution card; reuse in systematic-strategy desk engagement PH jurisdiction-aware promptingDTI/SEC/BIR/LGU rules, ₱3M VAT threshold, 8%-flat vs graduated election baked into prompts Encoded in skill prompts (not bolted on at the end) PROOF POINTERS ↗31 artifacts on disk (charters, briefs, ledgers) ↗Reuse pattern documented in systematic-strategy desk engagement methodology ## AI Systems Engineer (Independent) : Editorial & Content Automation Editorial / content automation·Remote Dec 2024 to Dec 2025 Built an 8-agent content-production pipeline and a 290-line AI-slop evaluator that drove a real draft from HEAVY (81) to CLEAN (3). Built an 8-agent content-production pipeline (topic-scout → researcher → drafter → editor → producer → art-director → chart-maker → exporter) and a 290-line AI-output (slop) evaluator scoring drafts on 13 literature-grounded metrics; drove a real draft from HEAVY (index 81) to CLEAN (index 3) . Engineered a dependency-free rendering pipeline (HTML/SVG → headless-Chrome PNG; Markdown → publish-ready PDF) so the output side of the platform runs without LLM API costs. Demonstrated that AI-assisted editorial work can be made measurable : each draft’s slop score, iteration delta, and shipped-iteration provenance are all logged. CONTRIBUTIONS · EVIDENCE · PROOF 8-agent content-production pipelinetopic-scout → researcher → drafter → editor → producer → art-director → chart-maker → exporter See /solutions/8-agent-editorial-pipeline 290-line AI-slop evaluator (13 metrics)Drove a real draft from HEAVY (81) to CLEAN (3) : 96% reduction OSS repo (slop-scanner project, /projects/slop-scanner) Dependency-free rendering pipelineHTML/SVG → headless-Chrome PNG; Markdown → publish-ready PDF; no LLM API cost on the output side Pipeline code on disk Measurable editorial workEach draft's slop score, iteration delta, and shipped-iteration provenance logged Logging + audit trail on disk PROOF POINTERS ↗Public slop-scanner repo with scorecard ↗Reuse pattern documented in systematic-strategy desk engagement methodology ## Trading Platform Testing & AI-Workflow Research Analyst (contract) AI-engineering consultancy (closed past contract; 04/2026 - 06/2026)·Remote Apr 2026 to Jun 2026 Structured testing of trading platform onboarding flows : execution clarity, workflow logic, and system usability from a trader-first perspective. Conducted structured testing of trading platform onboarding flows to evaluate execution clarity, workflow logic, and system usability from a trader-first perspective. Analyzed user execution pathways to identify breakdown points in order placement, setup processes, and trading workflow comprehension. Documented operational friction points and system inefficiencies affecting trading execution accuracy and user decision-making. Evaluated trading platform behavior under simulated real-user conditions to assess consistency, reliability, and functional clarity. Translated execution observations into structured insights to support product and trading workflow optimization. Applied AI prompting tools (ChatGPT, Claude, Gemini) to accelerate literature review and pattern recognition while manually verifying accuracy of all outputs. CONTRIBUTIONS · EVIDENCE · PROOF Onboarding-flow testingEnd-to-end onboarding as a fresh user; documented every friction point with priority-ordered remediation set 12-page remediation brief delivered to client Q3 scope Execution-clarity analysisReviewed order-execution copy, fee disclosures, risk warnings; flagged 14 ambiguities Documented in remediation brief AI-prompting for literature reviewAccelerated literature review and pattern recognition with manual verification of all outputs AI-assisted review notes delivered Trader-first usability insightIdentified breakdown points in order placement, setup processes, and trading workflow comprehension Operational-friction report delivered PROOF POINTERS ↗12-page remediation brief (delivered to client) ↗Operational-friction report (delivered) ## Crypto Trading Systems & Workflow Research Assistant Ledger51 Trading Community·Remote Oct 2025 to Apr 2026 Structured analysis of crypto trading workflows : order execution, bots, and platform mechanics. Supported structured analysis of crypto trading workflows , including order execution systems, trading bots, and platform mechanics. Simplified complex trading system behavior into structured, step-by-step execution frameworks for applied user understanding. Assisted in identifying workflow inefficiencies and execution gaps affecting trading accuracy and consistency. Provided analytical breakdowns of trading platform functions including order types, execution logic, and system interactions . Contributed to structured documentation of trading processes for research and training purposes. CONTRIBUTIONS · EVIDENCE · PROOF Workflow-pattern catalogMapped 8 community-trader workflows into typed-step process diagrams; basis for bot-development backlog Catalog on disk; used by team's bot backlog Order-execution comparison6 exchange stacks researched (Binance/Bybit/OKX/Kraken/Coinbase Pro/dYdX); latency, fee tiers, slippage, rate-limits Per-stack cheat sheets delivered Trading-bot pattern testingTested 4 community bot strategies on a sandbox exchange; documented mechanics + 7 failure modes Test reports delivered to community Perpetuals mechanics explainersFunding-rate dynamics, cross-margin vs isolated, liquidation cascades 3 explainer memos in community knowledge base PROOF POINTERS ↗Workflow catalog on disk ↗Per-stack cheat sheets delivered ↗Test reports on disk ↗3 explainer memos published ## Independent AI Systems Engineer : Personal Portfolio & Self-Directed Study Personal Portfolio & Self-Directed Study·Remote Jan 2025 to May 2026 Designed and shipped runnable AI projects end-to-end (RAG, ReAct, MCP, judge-harness, reflection loop, slop gate) and backtested systematic strategies. Designed and shipped multi-agent LLM systems end-to-end: agent charters, eval harnesses, model-routing policy, structured-output contracts, persistent memory, and Python tooling : for content automation, venture incubation, and personal-knowledge workflows. Designed and backtested systematic strategies on crypto and equities using Python; evaluated performance via Sharpe ratio, drawdown, win rate, and profit factor. Read and summarized 20+ academic and practitioner papers on momentum, mean-reversion, volatility carry, statistical arbitrage, and multi-agent AI architectures into structured one-page research notes. Built, audited, and open-sourced a portfolio of runnable AI projects : RAG scorecard, ReAct tool-calling agent with OTel traces, LLM-as-judge validated vs humans, MCP eval server, self-critiquing reflection agent, AI-slop evaluation gate. Maintained a research journal documenting hypotheses, methodology, statistical tests, and outcomes : building a personal library of structured AI + quant knowledge. CONTRIBUTIONS · EVIDENCE · PROOF 6 OSS multi-agent LLM projectsRAG, ReAct tool-calling, LLM-as-judge, MCP eval server, reflection loop, slop-scanner : each with scorecard See /projects/ (6 AI projects) 9 reproducible public-data quant projectsMultiple-testing corrections, locked OOS windows, look-ahead-bias audits See /projects/ (9 quant projects) 31-gate eval harness methodologyFull taxonomy on /methodology See /methodology AI Architecture research journalWorkbooks on context engineering, loop engineering, quant engineering with the M-series See /publications (in progress) 20+ research paper summariesMomentum, mean-reversion, volatility carry, statistical arbitrage, multi-agent AI architectures One-page research notes (in research journal) PROOF POINTERS ↗6 AI project repos with scorecards ↗9 quant project memos with figures ↗Methodology page with G1-G31 taxonomy ↗Workbook drafts available on request ## Financial Market Educator & AI Integration Specialist (ongoing side) Independent / Various Universities and Communities·Remote / Davao RegionCurrent Dec 2024 to Present Workshops, webinars, and guest lectures on financial markets, AI applications, and emerging technologies for students and professionals. Delivered workshops, webinars, and presentations on financial markets, trading systems, AI applications, and emerging technologies. Simplified complex financial and macroeconomic concepts into practical insights for students and professionals. Conducted market trend analysis involving forex, stocks, cryptocurrencies, and AI-driven financial systems . Utilized AI tools and data-driven research methodologies to improve educational delivery and strategic analysis. Served as a guest speaker at universities and professional communities on topics involving financial literacy, trading psychology, AI, and blockchain technologies. CONTRIBUTIONS · EVIDENCE · PROOF Workshops & webinars~110 attendees cumulative across PSHS-SMC, USeP CBA, Ledger51 community Workshop materials + recordings on request Guest lecturesUSeP CBA Annual Business Expo 2026 (AI in systematic trading); PSHS-SMC alumni series (AI engineering path) See /publications (press section) Financial-literacy curriculum6-module curriculum aligned with STA Tier-1 syllabus; piloted with 30 students in Davao Region Curriculum on request AI-in-finance articles3 Medium-published articles (operator-vs-owner, fintech-heavy-tells, most-expensive-thing-about-money) See /publications Trend analysisForex, stocks, cryptocurrencies, and AI-driven financial systems Market notes in research journal PROOF POINTERS ↗STA Tier-1 CTA certificate (program completed Dec 2025) ↗USeP CBA Annual Business Expo 2026 invitation (public talk) ↗PSHS-SMC alumni speaker invitation ↗3 Medium articles (publicly accessible) ## Guest Lecturer : Understanding the ICC and Indigenous Group in Mindanao University of Southeastern Philippines (USeP), College of Engineering·Davao Region, Philippines (in-person) Sep 2024 to Sep 2024 Delivered the Manobo / ICC / IPs guest lecture for USeP EGE 313, authored with a ChatGPT-assisted drafting pipeline that pre-figures the 31-gate evaluation harness now standard across my work. ## Live lecture Delivered to USeP EGE 313 (Understanding the ICC and Indigenous Group in Mindanao) for ECE 3A on the 2:30pm MW schedule, AY 2024-2025. 19-slide deck on the Manobo people of Mindanao, indigenous cultural communities (ICCs) and indigenous peoples (IPs), and the cultural-rights framework under the UNDRIP and the Philippine IPRA (RA 8371). The recording shows the PowerPoint Edit-mode title ( MACION. MANOBO. PIC. ECE3A. 230pm. MW ) and the live webcam feed in the lower-right corner : provenance for both the deck and the speaker. ## AI-assisted authoring pipeline Three weeks before the lecture, the slide deck was drafted end-to-end via a ChatGPT-assisted authoring pipeline. The recording below shows the AI-engineering workflow that produced the deck : the same precursor pattern that today runs through a 31-gate evaluation harness before any artifact ships. This is the lineage of the eval-first method: a 2024 single-agent drafting flow → the current 11-agent orchestrator + 31-gate harness. Public, NDA-clean, on-disk provenance preserved. CONTRIBUTIONS · EVIDENCE · PROOF Lecture deliveryLive Presenter-view delivery to USeP EGE 313 (ECE 3A, 2:30pm MW, AY 2024-2025) : 19-slide deck on Manobo / ICC / IPs / cultural rights Live recording on /proof AI-assisted authoring pipelineSlide deck drafted end-to-end via a ChatGPT-assisted drafting pipeline : same precursor pattern to the current 31-gate eval harness. Visual proof in the AI workflow recording. AI workflow recording on /proof Public reproducibilityBoth recordings (lecture + AI workflow) are public, NDA-clean, and show the file metadata proving provenance (PowerPoint title 'MACION. MANOBO. PIC. ECE3A. 230pm. MW'). On-disk provenance on /proof PROOF POINTERS ↗/proof/ [ALL 22 PROOFS →](/portfolio/proof/) --- # Skills · Christian T. Macion > Domain-grouped skill matrix with self-rated proficiency and years of practice: AI & multi-agent systems, quantitative research & statistics, and engineering tools. Source: `/skills/` SKILLS · 3 DOMAINS · ONE STACKSELF-RATED · LADDER 1 to 5 · YEARS OF PRACTICE CHRISTIAN.T.MACION·UTC+8·3 DOMAINS·SELF-RATED·OWNER-VERIFIED # 3 domains. One stack. Skills grouped the way a hiring manager reads them: what I build, how I validate it, and what I ship it with. Each skill is paired with a self-rated proficiency ladder (1 = exposure ·5 = production) and the number of years I've practiced it. Self-rated, not self-inflated. Proficiency dots are my honest answer to "could I ship this to production on Monday without help?" Tradeoffs between languages, libraries, and domains are explicit: e.g., I can ship production Python but my Docker is working , not expert . Years of practice reflect actual use, not calendar time since first exposure.stack=build⋅validate⋅shipwherevalidate=G1..G31(1)domains∣D∣3threeskills∣S∣36in stackratedrated36of 36 (100%)★ production∣s:ℓ(s)=5∣8level-5 ratedProficiency ladder:1exposure2familiar3working4strong5production⏱yrspractice (self-reported) 00 / skills Domain 1 ## AI & Multi-Agent Systems How I build, evaluate, and ship LLM systems that hold up in production. ### Multi-Agent Orchestration 5/5 orchestrator-worker / supervisor-worker topologies ⏱3yrs ### LLM Evaluation Harness Design 5/5 block-bootstrap CIs · random-timing nulls · position-bias flips ⏱2yrs ### Model Routing & Token-Cost Optimization 4/5 Opus = judgment · Sonnet = assembly · Haiku = mechanical · Fable = hardest autonomous ⏱2yrs ### Context Engineering 5/5 agent charters · hand-off packets · few-shot routing ⏱3yrs ### Structured-Output Contracts & Validation 5/5 JSON schema · mechanical validator (must exit 0) ⏱3yrs ### Tool Use / Function Calling 4/5 ReAct · loop guards · fault injection ⏱2yrs ### RAG / Chunking / Source-Grounded Generation 4/5 retrieval scorecard · recall@k · MRR · faithfulness ⏱2yrs ### Prompt Engineering & Caching 4/5 few-shot · CoT · system-prompt discipline ⏱3yrs ### Persistent Agent Memory & MCP 4/5 file-based cross-linked records · MCP servers · conformance tests ⏱1yr ### OpenTelemetry for LLM Apps 3/5 gen_ai.* attributes · per-step spans ⏱1yr ### LLM-as-Judge Validation 4/5 Cohen's κ vs human raters · bootstrap CI · bias measurement ⏱2yrs ### AI-Slop / Content-Quality Gates 4/5 13 literature-grounded metrics · stdlib scoring core ⏱2yrs Domain 2 ## Quantitative Research & Statistics The research-process disciplines a quant-researcher seat is hired for. ### Statistical & Mathematical Modeling 4/5 linear algebra · probability · time-series ⏱4yrs ### Pre-Registration & Multiple-Testing Control 4/5 Deflated Sharpe · PBO (CSCV) · MinBTL ⏱2yrs ### Backtesting & Performance Metrics 5/5 Sharpe · Sortino · drawdown · win-rate · profit factor ⏱3yrs ### Walk-Forward & Out-of-Sample Validation 4/5 frozen-spec · pre-registered windows · paper-shadow ⏱2yrs ### Block-Bootstrap & Monte-Carlo 4/5 preserves autocorrelation · null generation · CI construction ⏱2yrs ### Transaction-Cost Modeling 3/5 spread · slippage · linear impact · break-even turnaround ⏱2yrs ### Look-Ahead Bias Control 4/5 one-bar shift test · point-in-time data discipline ⏱2yrs ### Volatility Modeling 3/5 realized vol · GARCH-class · variance risk premium ⏱2yrs ### Cointegration & Pairs Trading 3/5 ADF · half-life · mean-reversion diagnostics ⏱1yr ### Cross-Sectional & Time-Series Alpha 4/5 momentum · carry · mean-reversion · regime overlays ⏱2yrs ### Crypto & Derivatives Microstructure 4/5 perpetual funding · basis · term structure ⏱2yrs ### Risk Management & Position Sizing 4/5 vol targeting · regime-aware exposure · DD-aware scaling ⏱2yrs Domain 3 ## Engineering & Tools The day-to-day stack that ships the work above. ### Python 5/5 numpy · pandas · matplotlib · pyarrow · boto3 · scikit-learn ⏱4yrs ### SQL 4/5 SQLite · Postgres · point-in-time queries ⏱4yrs ### AWS S3 4/5 data lake · idempotent incremental pulls · byte-range subsetting ⏱2yrs ### Databento 3/5 intraday market data · historical archives ⏱1yr ### Docker 4/5 working : multi-agent + MCP service containers ⏱2yrs ### TradingView + Pine Script 4/5 indicator prototyping · strategy backtesting ⏱2yrs ### Google Colab / Jupyter 5/5 exploratory research · reproducible notebooks ⏱4yrs ### Git & GitHub 4/5 CI/CD · Pages deploy · public reproducibility ⏱4yrs ### Claude Agent SDK / MCP 4/5 MCP servers · Tools/Resources/Prompts primitives ⏱1yr ### Headless-Chrome Rendering 4/5 HTML/SVG → PNG · Markdown → PDF pipelines ⏱2yrs ### Microsoft Excel / Google Sheets 5/5 financial modeling · FP&A · collaborative analysis ⏱6yrs ### Notion 4/5 knowledge base · agent-memory governance ⏱3yrs --- # Certifications · Christian T. Macion > Curated 49+ of 102 professional certifications, grouped by domain and Tier-1 issuer. Source: `/certifications/` 01. certifications CERTIFICATIONS · 102 TOTAL · 4 DOMAINSISSUER-VERIFIED · TIER-1 FLAGGED · CURATED BELOW CHRISTIAN.T.MACION·UTC+8·102 CERTS·4 DOMAINS·OWNER-VERIFIED # 49+ of 102 certifications, 4 domains. Every cert links to its issuer's public registry. or it doesn't ship on this page. 102 professional certifications across AI, finance, math/statistics, and event flagships to collected over 48 months (2023 → 2026). Curated below by domain; Tier-1 issuers flagged. cert_density=tc(t)​=48 months102 certs​≈2.1 certs/monthcurated49belowtotal102↑all-timeflagships3NASA · Meta · IBM · IMCtier-123STA · SEC · BSP Domain 1 ## AI, Data & Technology Flagship AI and data-science credentials across Tier-1 issuers (IBM, Google, Cisco, DataCamp, Ateneo de Davao). https://www.ateneo.edu/AI for the Modern Workforce: From Blockchain to the MOVE Language : Ateneo de Davao University + US Embassy American Spaces Philippines, Nov 8 2025. Co-signed by Lorgina Samson (Director, University Libraries) and Kevin Punzalan (American Spaces Philippines Specialist, US Embassy in the Philippines). Artificial Intelligence FundamentalsIBM2026Tier 1 AI Fundamentals with IBM SkillsBuildCisco2025 AI FundamentalsDataCamp2025 Introduction to AIGoogle2025Tier 1 Introduction to AI AgentsDataCamp2026Tier 1 AI for the Modern Workforce: From Blockchain to the MOVE LanguageAteneo de Davao University2025Tier 1 Introduction to Modern AICisco Networking Academy2025 Introduction to Data ScienceCisco2024 Python Essentials 1Cisco2024 Software Development and Design ThinkingDICT2024Tier 1 Apply AI: Analyze Customer ReviewsCisco2025 Blockchain4Youth (B4Y-2026-000701)Bitget2026Tier 1 Domain 2 ## Finance, Trading & Economics Tier-1 finance credentials: STA flagship + Goldman / JPMorganChase brand-name work + DataCamp finance stack. https://www.sta-uk.com/Tier-1 Certified Technical Analyst® (CTA) · Society of Technical Analysts · certificate #260197 · Jan 2026. Photographic proof of the designation. Certified Technical Analyst ProgramSociety of Technical Analysts of the Philippines (Tier-1)2025Flagship Certified Technical AnalystCertifyMe2025 Completed Goldman Sachs Asset Management educational module: Foundations of Growth Equity (GS Alternatives University, 2025)Goldman Sachs2025Tier 1 JPMorganChase : Investment Banking Job SimulationForage2026Tier 1 Financial Trading in PythonDataCamp2025Tier 1 Math for Finance ProfessionalsDataCamp2025Tier 1 Intermediate Python for FinanceDataCamp2025 SQL for Finance ProfessionalsDataCamp2025 Financial Modeling and Forecasting Financial StatementsLinkedIn2024 Financial Forecasting with Analytics Essential TrainingLinkedIn2024 Using Data in Financial AnalysisLinkedIn2024 Excel for Financial Planning and Analysis (FP&A)LinkedIn2024 Economics MasterclassupGrad2024 Principles of Economics : MacroeconomicsMarginal Revolution University2024 Domain 3 ## Math, Statistics & Analytics Math and analytics depth : statistics, data literacy, and the modeling foundations under the quant work. Probability & Statistics FoundationsDataCamp / Khan Academy / university curriculum2024 Data LiteracyDataCamp2025 Marketing Analytics for BusinessDataCamp2024 Fundamentals of AccountingSEC Philippines2024Tier 1 Introduction to Capital MarketSEC Philippines2024Tier 1 Financial ReportingSEC Philippines2024Tier 1 Introduction to CorporationSEC Philippines2024Tier 1 Financial PlanningBangko Sentral ng Pilipinas2023Tier 1 Digital Financial LiteracyBangko Sentral ng Pilipinas2023Tier 1 Operational Analysis of Suspicious Transaction ReportsBasel Institute on Governance2024Tier 1 Domain 4 ## Events, Hackathons & Distinguishing Flagships Distinguishing Tier-1 flagships that anchor a memorable profile: NASA Space Apps, Meta BIDA, university teaching, and Philippine civil-service eligibility. https://bida.gov.ph/BIDA × Bayan Academy × Meta AIccelerate 2025 · Nov 12-21 2025 (5-day hybrid training) · awarded Dec 17 2025. Co-signed by BIDA, Bayan Academy, and Meta Philippines. Galactic Problem Solver : NASA Space Apps Challenge (Zurich, CH)NASA2025Flagship BIDA META AICCELERATE 2025Meta2025Flagship Guest Speaker : CBA Annual Business Expo 2026University of Southeastern Philippines (USeP)2026Tier 1 Career Service Professional EligibilityCivil Service Commission (Philippines)2023Tier 1 Effective LeadershipHP LIFE2024 Agile Project ManagementHP LIFE2024 Introduction to IT Project ManagementUniversity of the Philippines2024Tier 1 Customer Experience (CX) for Business SuccessHP LIFE2024 AI for Business ProfessionalsHP LIFE2025 Diploma in CryptocurrencyAlison2024 Crypto Trading Deep DiveBinance Academy2024 Introduction to Cryptocurrency RecoveryAnChain.AI2025 Fundamentals of BlockchainAnChain.AI2025 [ALL 102 CERTS →](/portfolio/resume/) Full list: the curated 49 above is the highlight reel. The complete list of 102 (issuer + date + verifiable credential ID where applicable) lives in the[tailored resume package](/portfolio/resume/). Government certs include CSC Career Service Professional Eligibility, BSP financial planning, SEC Philippines capital market and financial reporting, and Basel Institute on Governance STR analysis. --- # Publications · Christian T. Macion > Public work. research projects, workbooks, talks, OSS repos, and press. All NDA-safe. Source: `/publications/` PUBLICATIONS · 29 ARTIFACTSDATED · LINKABLE · PROOF-IN-ROW # Things I have written, built, and said in public. Dated. Linkable. If a row says I did something, the proof is in the same row. Every artifact below is either an OSS repo, a memo, a workbook, or a public talk. all 29papers9oss 6workbooks4press 10 CHRISTIAN.T.MACION·UTC+8·29 ARTIFACTS·NDA-SAFE·OWNER-VERIFIED trust/[testednumerical-eval](/portfolio/proof//#code-artifacts)[reproduciblepublic repos](/portfolio/projects/)[NDA-cleanby construction](/portfolio/experience//#nda-safe)[RSS/feed.xml](/portfolio/feed.xml)verified2026-08-09 7 min read · last updated2026-08-09 · [how verified ↗](/portfolio/proof/) 00 / publications Research papers · 9 quant projects ## Public-data systematic research. ### 2026 #### [The Variance Risk Premium (VIX vs Realized)](/portfolio/projects/quant/04-variance-risk-premium/) Implied > realized 85% of 36 yrs; predicts returns, Newey-West t = +6.5. Years VRP > 085% Newey-West t-stat+6.5 Sample window1990 - 2025 Python (numpy / pandas / matplotlib)VIX (CBOE) historicalRealized-vol estimatorNewey-West HAC inferenceJan 2026[Read →](/portfolio/projects/quant/04-variance-risk-premium/) #### [Pairs Trading via Cointegration (BTC / ETH)](/portfolio/projects/quant/05-pairs-cointegration/) ADF −2.11 (not cointegrated), half-life 208 d : the fade loses, as the test predicts. ADF test statistic−2.11 Half-life (estimated)208 days OOS Sharpe (fade)≈ 0 Python (numpy / pandas / matplotlib)Augmented Dickey-Fuller (from-scratch)Half-life estimator5 bps / side cost modelFeb 2026[Read →](/portfolio/projects/quant/05-pairs-cointegration/) #### [Crypto Funding-Carry](/portfolio/projects/quant/06-funding-carry/) Funding +11.9% annualized premium; fade IS 1.11 → OOS −0.05 (decayed post-2023). Annualized funding+11.9% Fade IS Sharpe1.11 Fade OOS Sharpe−0.05 Python (numpy / pandas / matplotlib)Perpetual funding rates (Binance)Bootstrap CIPre-registered lock windowApr 2026[Read →](/portfolio/projects/quant/06-funding-carry/) #### [Macro / Volatility-Regime Overlay](/portfolio/projects/quant/07-macro-regime-overlay/) Vol-managed exposure: Sharpe 0.64 → 0.67, max DD −58% → −42%. Sharpe (base)0.64 Sharpe (regime overlay)0.67 Max DD (base → overlay)−58% → −42% Python (numpy / pandas / matplotlib)VIX regime classifierExposure scaler (smoothed)5 bps / side costMay 2026[Read →](/portfolio/projects/quant/07-macro-regime-overlay/) #### [Backtest Engine + Cost Model](/portfolio/projects/quant/08-backtest-engine-costs/) High-turnover signal wins gross (0.20 > 0.13) but loses net of cost : break-even 20 bps. Sharpe (gross)0.20 Sharpe (net, realistic)< 0 Break-even cost~20 bps / RT Python (numpy / pandas / matplotlib)Event-driven backtest loopLinear-impact cost modelBreak-even turnaround scanMay 2026[Read →](/portfolio/projects/quant/08-backtest-engine-costs/) #### [Look-Ahead Bias Audit (the shift test)](/portfolio/projects/quant/09-lookahead-bias-audit/) A leak inflates Sharpe 0.59 → 5.07; the shift test exposes it as 88% phantom. Clean Sharpe0.59 Leaked Sharpe5.07 Phantom % detected88% Python (numpy / pandas / matplotlib)One-bar shift testLeak-magnitude quantificationJun 2026[Read →](/portfolio/projects/quant/09-lookahead-bias-audit/) ### 2025 #### [Multiple Testing & the Deflated Sharpe Ratio](/portfolio/projects/quant/01-deflated-sharpe/) Best-of-160 BTC rule: IS Sharpe 1.14 is only 1.24× the pure-noise expectation : DSR = 0.70 (fail). Best-of-160 in-sample Sharpe1.14 Pure-noise expectation0.92 Deflated Sharpe Ratio0.70 Python (numpy / pandas / matplotlib)Block-bootstrap nullDeflated Sharpe Ratio160-rule BTC sweepSep 2025[Read →](/portfolio/projects/quant/01-deflated-sharpe/) #### [Cross-Sectional Momentum (18 coins)](/portfolio/projects/quant/02-cross-sectional-momentum/) IS Sharpe 0.91 → OOS −0.03 : an honest decay; bootstrap CI straddles zero. In-sample Sharpe0.91 Out-of-sample Sharpe−0.03 OOS 95% CIstraddles 0 Python (numpy / pandas / matplotlib)Cross-sectional rankingBootstrap confidence intervalLocked out-of-sample windowOct 2025[Read →](/portfolio/projects/quant/02-cross-sectional-momentum/) #### [Time-Series Momentum + Vol Targeting](/portfolio/projects/quant/03-timeseries-momentum-voltarget/) Vol-targeting lifts Sharpe 0.27 → 0.39 and halves max DD (−62% → −30%). Sharpe (raw)0.27 Sharpe (vol-targeted)0.39 Max DD (raw → targeted)−62% → −30% Python (numpy / pandas / matplotlib)Time-series momentum signal20-day realized-vol targeting5 bps / side cost modelDec 2025[Read →](/portfolio/projects/quant/03-timeseries-momentum-voltarget/) OSS projects · 6 AI builds ## Multi-agent LLM systems with eval-first discipline. ### 2026 #### [LLM-as-Judge Harness](/portfolio/projects/ai/03-judge-harness/) LLM-as-judge pipeline validated against human raters : Cohen's κ = 0.58 with bootstrap CI and position-bias measured. Cohen's κ (vs human)0.58 Pass rate0.60 ± 0.02 Position bias17% PythonLLM-as-judge prompt patternCohen's κ + bootstrap CIPosition-bias flip testJan 2026[Read →](/portfolio/projects/ai/03-judge-harness/)[Repo ↗](https://github.com/christianmacion/judge-harness) #### [Eval MCP Server](/portfolio/projects/ai/04-eval-mcp-server/) MCP server exposing the slop-evaluation gate over Tools, Resources, and Prompts : 20/20 conformance, 100% round-trip parity. MCP conformance20 / 20 Round-trip parity100% Primitives exposed3 / 3 PythonModel Context Protocol (MCP)Tools / Resources / Prompts primitives13-metric slop gateFeb 2026[Read →](/portfolio/projects/ai/04-eval-mcp-server/)[Repo ↗](https://github.com/christianmacion/eval-mcp-server) #### [Reflect-Revise Loop](/portfolio/projects/ai/05-reflect-revise/) Reflection-loop agent : mean SLOP 127.5 → 14.0 across drafts, 3/4 improved, 1/4 honest no-progress halt. Mean SLOP score127.5 → 14.0 Drafts improved3 / 4 Honest no-progress halts1 / 4 PythonReflect / revise loopHonest no-progress haltPer-iteration score-delta loggingMar 2026[Read →](/portfolio/projects/ai/05-reflect-revise/)[Repo ↗](https://github.com/christianmacion/reflect-revise) #### [Slop Scanner](/portfolio/projects/ai/06-slop-scanner/) 13-metric literature-grounded AI-output quality gate : drove a real draft from HEAVY (81) to CLEAN (3). Real-draft improvement81 → 3 Metrics13 External dependencies0 Python (stdlib)Streamlit13 literature-grounded metricsMar 2026[Read →](/portfolio/projects/ai/06-slop-scanner/)[Repo ↗](https://github.com/christianmacion/slop-scanner) ### 2025 #### [RAG Recall Eval](/portfolio/projects/ai/01-rag-recall/) RAG service that proves its own retrieval : recall@3 = 0.886, MRR@3 = 0.805 with offline stdlib TF-IDF retriever. recall@30.886 MRR@30.805 Mean faithfulness1.00 Hallucination flags0 / 35 Python (stdlib only : no LLM API, no pip install)TF-IDF retriever17-doc / 34-chunk corpusStreamlit appNov 2025[Read →](/portfolio/projects/ai/01-rag-recall/)[Repo ↗](https://github.com/christianmacion/rag-recall) #### [Tool-Call Agent](/portfolio/projects/ai/02-toolcall-agent/) ReAct-style tool-calling agent with OTel traces, fault injection, and 100% tool/arg correctness. Tool / arg correctness100% Injected faults recovered6 / 6 Loops without bound0 PythonReAct loopOpenTelemetry (gen_ai.* attrs)Fault-injection harnessDec 2025[Read →](/portfolio/projects/ai/02-toolcall-agent/)[Repo ↗](https://github.com/christianmacion/toolcall-agent) Workbooks · long-form ## 1 published · 3 in progress. Long-form workbooks that consolidate method, code, and exercises. Published entries ship as public PDFs; in-progress drafts are available on request. ### [Kaufman Efficiency Ratio. Trading Research Note](/portfolio/workbooks/kaufman-efficiency-ratio.pdf) Adaptive efficiency measurement for trading strategies; reproducible Python + backtest manifest.4 pp·trading researchers, strategy developers·Christian T. MacionPublished : [287 KB · 4 pp](/portfolio/workbooks/kaufman-efficiency-ratio.pdf) · 2026-07-18 · cite:macion2026kaufman ### AI Architecture & Multi-Agent Systems Workbook 11-agent orchestrator, eval harness, MCP server, model routing~80 pp·AI engineers, agent architects, eval harness designersIn progress to early-reader draft; full PDF available on request (non-public, gated by topic) ### Statistical Modeling for Systematic Trading Workbook multiple testing, walk-forward, block-bootstrap, deflated Sharpe, cost realism~110 pp·quant researchers, strategy developers, backtest engineersIn progress to early-reader draft; full PDF available on request (non-public, gated by topic) ### STA Tier-1 CTA Curriculum Workbook technical analysis fundamentals through advanced pattern recognition~150 pp·technical analysts, finance students, CFA/CMT candidatesIn progress to early-reader draft; full PDF available on request (non-public, gated by topic) Press & recognition ## Public talks, lectures, and recognitions. 2026 ### Frontier Models for Quant Research & AI Engineering. r3 consulting brief released (18 pp, citable) Venue: portfolio.os · /research/frontier-models Public-shareable brief comparing M3, Kimi K3, and Claude Opus 4.8 (with Fable 5, GPT-5.6 Sol, Gemini 3 Pro as reference points) for quant research + AI engineering. 18-page letter PDF, footnoted sources, two verification gates, declared gaps. 2026 ### v6.0.5 site patch. three real bugs shipped (Atom feed base path, AI H1 slugs, methodology cross-linking) Venue: portfolio.os · commit history Three post-deploy audit fixes shipped: (1) feed.xml/feed-projects.xml/feed-solutions.xml entry URLs now include the /portfolio base path; (2) AI research page H1s now show human titles (RAG Recall Eval, Tool-Call Agent, LLM-as-Judge Harness, Eval MCP Server, Reflect-Revise Loop, Slop Scanner) instead of filename slugs; (3) /methodology now cross-links 12 unique glossary terms (DSR, PBO, walk-forward, block-bootstrap, embargo, OOS, slippage, Sharpe, regime, Bonferroni-Holm, MinBTL, survivorship) into the /glossary/{id} deep pages, up from zero. 2026 ### v6.0.4 spine shipped. 25 glossary deep pages + 15 research citation pages Venue: portfolio.os · release notes Two new content surfaces: /glossary/{slug} (25 deep pages, one per glossary term, with extended prose + 3-6 related-term cross-links each) and /research/{lane}/{0N-slug} (15 academic citation pages, one per quant + AI project, where the numeric prefix comes from MDX filename order; URLs rendered as /research/quant/01-deflated-sharpe-ratio etc., with auto-generated BibTeX entries + copy-to-clipboard button + canonical-link row to /projects). Total 80 pages indexed. FlagshipCard self-link bug fixed (3 certs now point to issuer URLs). All build-time dates switched to UTC+8. 2026 ### Vol-target time-series momentum paper-trade. live publication on /positions Venue: portfolio.os · /positions Live paper-trade started on project 03 (BTC/USDT, 30% annualized vol target). Public-data reproducible end-to-end. Currently in regime 2 (50+ trading days), tracking slippage against the 5 bps linear-impact model. 4.2 bps realized. 2026 ### Slop-scanner app released. 13-metric literature-grounded AI-output quality gate Venue: github.com/christianmacion26/slop-scanner Streamlit app that scores a draft on 13 literature-grounded quality metrics (vocabulary entropy, hedge-word density, generic-template phrases, repetition n-grams) and outputs a single SLOP index from HEAVY to CLEAN. Stdlib scoring core needs no LLM and no API key. Drove a real draft from HEAVY (81) to CLEAN (3). 2026 ### CTA program graduate. Society of Technical Analysts of the Philippines Venue: STA Philippines Cert #260197 issued Jan 19 2026 (program completed Dec 2025; cert-grant ceremony Jan 19 2026). Photographic proof on the record. same image used in the hero signature strip. 2025 ### STA Tier-1 CTA Program completion. Society of Technical Analysts of the Philippines Venue: STA Philippines Tier-1 Certified Technical Analyst program, completed Dec 2025. Final capstone: regime-classification case study on ASEAN equities. 2025 ### BIDA × Meta AIccelerate 2025. 5-day hybrid training, awarded Dec 17 2025 Venue: BIDA × Bayan Academy × Meta Selected for the BIDA × Bayan Academy × Meta AIccelerate 2025 Batch 3 (Nov 12 to 21 2025). Five-day hybrid training on applied AI for MSMEs; co-signed certificate from BIDA, Bayan Academy, and Meta Philippines. 2025 ### University guest lecturer. University of Southeastern Philippines (USeP) Venue: USeP College of Engineering Guest lecture on systematic trading research and AI for engineering students. Course: CpE / EE 5th-year elective. 2024 ### PSHS-SMC alumni speaker. Philippine Science High School Southern Mindanao Venue: PSHS-SMC Alumni Talk Series Talk on AI engineering and the educational path from PSHS to AI/quant careers. ~80 attendees (alumni + current students). ## Want a specific paper or talk? Email me the topic and I'll point you to the artifact (PDF, video, repo) within 24 hours. [Email me](/portfolio/resume/)[See methodology](/portfolio/methodology/) --- # Projects · Christian T. Macion > AI engineering projects (multi-agent, eval, MCP) and quantitative research projects (multiple-testing, cross-sectional alpha, volatility, cointegration). Source: `/projects/` 01. projects Portfolio # 15 projects. Two lanes. Metrics real. Methods documented. Code on disk. Every metric is real, every method is documented, every line of code is on disk. Filter by lane to focus on the role you're hiring for. 15projects shipped6AI engineer projects9quant research projects15open-source repos[All 15](/portfolio/projects/)[AI Engineer 6](/portfolio/projects/?tag=ai)[Quant Researcher 9](/portfolio/projects/?tag=quant)UPDATES10:32AVAILQ3 2026 · OPEN TO REMOTE QR / AI ROLES · <24H REPLY·10:31SHIPv7.4 MOTION LAYER · SITE-WIDE · 8 PRIMITIVES·10:28PROJ15 PROJECTS · 9 QUANT · 6 AI · 2 LANES·10:25PAPERTSM-VT · +0.42% · σ 1.1 · WEEK 27·10:22READSTOCHASTIC CALCULUS · SHREVE CH.4 · ITÔ LEMMA·10:18BUILD76.5K LOC PYTHON · 31 GATES · ZERO ROLLBACKS·10:14GATEG31 · LOOKAHEAD AUDIT · 88% CATCH ON PHANTOM LEAK·10:09ALPHACOINTEGRATION PAIRS · 5Y HALF-LIFE · BLOCK-BOOTSTRAP·10:05STATUSDUTY · ONLINE · 24H REPLY · NDA-CLEAN·10:32AVAILQ3 2026 · OPEN TO REMOTE QR / AI ROLES · <24H REPLY·10:31SHIPv7.4 MOTION LAYER · SITE-WIDE · 8 PRIMITIVES·10:28PROJ15 PROJECTS · 9 QUANT · 6 AI · 2 LANES·10:25PAPERTSM-VT · +0.42% · σ 1.1 · WEEK 27·10:22READSTOCHASTIC CALCULUS · SHREVE CH.4 · ITÔ LEMMA·10:18BUILD76.5K LOC PYTHON · 31 GATES · ZERO ROLLBACKS·10:14GATEG31 · LOOKAHEAD AUDIT · 88% CATCH ON PHANTOM LEAK·10:09ALPHACOINTEGRATION PAIRS · 5Y HALF-LIFE · BLOCK-BOOTSTRAP·10:05STATUSDUTY · ONLINE · 24H REPLY · NDA-CLEAN·METHODEVAL-FIRST·31 GATES · ZERO ROLLBACKSNDA-CLEAN·PUBLIC-DATA REPRODUCIBLEALPHA-DRIVEN·14 SIGNALS · 9 STRATEGIES · 1 METHODSENIOR·QUANT RESEARCHER · AI ENGINEERDETERMINISTIC·BUILD-SEEDED · NO RANDOM AT RENDEROPS-RESEARCH·MULTI-AGENT · EVAL-DRIVEN · NDA-CLEANMULTI-AGENT·11 AGENTS · 1 ORCHESTRATORSHIP-READY·GATES PASS · DEPLOY MIRROROPEN TO WORK·QR / AI · REMOTE · UTC+8PHILIPPINES·DIGOS CITY · DAVAO DEL SUREVAL-FIRST·31 GATES · ZERO ROLLBACKSNDA-CLEAN·PUBLIC-DATA REPRODUCIBLEALPHA-DRIVEN·14 SIGNALS · 9 STRATEGIES · 1 METHODSENIOR·QUANT RESEARCHER · AI ENGINEERDETERMINISTIC·BUILD-SEEDED · NO RANDOM AT RENDEROPS-RESEARCH·MULTI-AGENT · EVAL-DRIVEN · NDA-CLEANMULTI-AGENT·11 AGENTS · 1 ORCHESTRATORSHIP-READY·GATES PASS · DEPLOY MIRROROPEN TO WORK·QR / AI · REMOTE · UTC+8PHILIPPINES·DIGOS CITY · DAVAO DEL SURproject updates · seeded from BUILD_DATEEVAL-FIRST·31 GATES · ZERO ROLLBACKSNDA-CLEAN·PUBLIC-DATA REPRODUCIBLEALPHA-DRIVEN·14 SIGNALS · 9 STRATEGIESMULTI-AGENT·6 AI PROJECTS · 1 ORCHESTRATORCOINTEGRATION·PAIRS · 5Y HALF-LIFEVARIANCE RISK·PREMIUM HARVEST · DELTA-HEDGEDREGIME OVERLAY·VOL-TARGETED · DEFENSIVECRYPTO NATIVE·BTC · ETH · FUNDING-RATE CARRYLOOKAHEAD AUDIT·88% CATCH ON PHANTOM LEAKSHIP-READY·GATES PASS · DEPLOY MIRROREVAL-FIRST·31 GATES · ZERO ROLLBACKSNDA-CLEAN·PUBLIC-DATA REPRODUCIBLEALPHA-DRIVEN·14 SIGNALS · 9 STRATEGIESMULTI-AGENT·6 AI PROJECTS · 1 ORCHESTRATORCOINTEGRATION·PAIRS · 5Y HALF-LIFEVARIANCE RISK·PREMIUM HARVEST · DELTA-HEDGEDREGIME OVERLAY·VOL-TARGETED · DEFENSIVECRYPTO NATIVE·BTC · ETH · FUNDING-RATE CARRYLOOKAHEAD AUDIT·88% CATCH ON PHANTOM LEAKSHIP-READY·GATES PASS · DEPLOY MIRRORproject discipline · 10 anchor frames Lane 1 · AI Engineer ## Multi-agent LLM systems & eval AI#01open source ### [RAG Recall Eval](/portfolio/projects/ai/01-rag-recall/) RAG service that proves its own retrieval : recall@3 = 0.886, MRR@3 = 0.805 with offline stdlib TF-IDF retriever. recall@30.886over 35 labeled Q'sMRR@30.805Mean faithfulness1.00[Read more →](/portfolio/projects/ai/01-rag-recall/)AI#02open source ### [Tool-Call Agent](/portfolio/projects/ai/02-toolcall-agent/) ReAct-style tool-calling agent with OTel traces, fault injection, and 100% tool/arg correctness. Tool / arg correctness100%Injected faults recovered6 / 6Loops without bound0[Read more →](/portfolio/projects/ai/02-toolcall-agent/)AI#03open source ### [LLM-as-Judge Harness](/portfolio/projects/ai/03-judge-harness/) LLM-as-judge pipeline validated against human raters : Cohen's κ = 0.58 with bootstrap CI and position-bias measured. Cohen's κ (vs human)0.58moderate agreementPass rate0.60 ± 0.0295% bootstrap CIPosition bias17%lower = better; flip-sensitive[Read more →](/portfolio/projects/ai/03-judge-harness/)AI#04open source ### [Eval MCP Server](/portfolio/projects/ai/04-eval-mcp-server/) MCP server exposing the slop-evaluation gate over Tools, Resources, and Prompts : 20/20 conformance, 100% round-trip parity. MCP conformance20 / 20across all 3 primitivesRound-trip parity100%Primitives exposed3 / 3Tools · Resources · Prompts[Read more →](/portfolio/projects/ai/04-eval-mcp-server/)AI#05open source ### [Reflect-Revise Loop](/portfolio/projects/ai/05-reflect-revise/) Reflection-loop agent : mean SLOP 127.5 → 14.0 across drafts, 3/4 improved, 1/4 honest no-progress halt. Mean SLOP score127.5 → 14.0across draftsDrafts improved3 / 4Honest no-progress halts1 / 4shipped original[Read more →](/portfolio/projects/ai/05-reflect-revise/)AI#06open source ### [Slop Scanner](/portfolio/projects/ai/06-slop-scanner/) 13-metric literature-grounded AI-output quality gate : drove a real draft from HEAVY (81) to CLEAN (3). Real-draft improvement81 → 3HEAVY → CLEANMetrics13External dependencies0[Read more →](/portfolio/projects/ai/06-slop-scanner/) Lane 2 · Quantitative Researcher ## Systematic strategies & statistical validation Quant#01public-data reproducible ### [Multiple Testing & the Deflated Sharpe Ratio](/portfolio/projects/quant/01-deflated-sharpe/) Best-of-160 BTC rule: IS Sharpe 1.14 is only 1.24× the pure-noise expectation : DSR = 0.70 (fail). Best-of-160 in-sample Sharpe1.14Pure-noise expectation0.92Deflated Sharpe Ratio0.70fails the multiple-testing adjustment[Read more →](/portfolio/projects/quant/01-deflated-sharpe/)Quant#02public-data reproducible ### [Cross-Sectional Momentum (18 coins)](/portfolio/projects/quant/02-cross-sectional-momentum/) IS Sharpe 0.91 → OOS −0.03 : an honest decay; bootstrap CI straddles zero. In-sample Sharpe0.91Out-of-sample Sharpe−0.03honest decayOOS 95% CIstraddles 0cannot reject the null[Read more →](/portfolio/projects/quant/02-cross-sectional-momentum/)Quant#03public-data reproducible ### [Time-Series Momentum + Vol Targeting](/portfolio/projects/quant/03-timeseries-momentum-voltarget/) Vol-targeting lifts Sharpe 0.27 → 0.39 and halves max DD (−62% → −30%). Sharpe (raw)0.27Sharpe (vol-targeted)0.39Max DD (raw → targeted)−62% → −30%[Read more →](/portfolio/projects/quant/03-timeseries-momentum-voltarget/)Quant#04public-data reproducible ### [The Variance Risk Premium (VIX vs Realized)](/portfolio/projects/quant/04-variance-risk-premium/) Implied > realized 85% of 36 yrs; predicts returns, Newey-West t = +6.5. Years VRP > 085%of 36 yrsNewey-West t-stat+6.5HAC-consistentSample window1990 - 2025[Read more →](/portfolio/projects/quant/04-variance-risk-premium/)Quant#05public-data reproducible ### [Pairs Trading via Cointegration (BTC / ETH)](/portfolio/projects/quant/05-pairs-cointegration/) ADF −2.11 (not cointegrated), half-life 208 d : the fade loses, as the test predicts. ADF test statistic−2.11fail to reject unit root at 5%Half-life (estimated)208 daysOOS Sharpe (fade)≈ 0as the test predicted[Read more →](/portfolio/projects/quant/05-pairs-cointegration/)Quant#06public-data reproducible ### [Crypto Funding-Carry](/portfolio/projects/quant/06-funding-carry/) Funding +11.9% annualized premium; fade IS 1.11 → OOS −0.05 (decayed post-2023). Annualized funding+11.9%BTC perp, ~30-day centered meanFade IS Sharpe1.11Fade OOS Sharpe−0.05decay post-2023[Read more →](/portfolio/projects/quant/06-funding-carry/)Quant#07public-data reproducible ### [Macro / Volatility-Regime Overlay](/portfolio/projects/quant/07-macro-regime-overlay/) Vol-managed exposure: Sharpe 0.64 → 0.67, max DD −58% → −42%. Sharpe (base)0.64Sharpe (regime overlay)0.67Max DD (base → overlay)−58% → −42%[Read more →](/portfolio/projects/quant/07-macro-regime-overlay/)Quant#08public-data reproducible ### [Backtest Engine + Cost Model](/portfolio/projects/quant/08-backtest-engine-costs/) High-turnover signal wins gross (0.20 > 0.13) but loses net of cost : break-even 20 bps. Sharpe (gross)0.20Sharpe (net, realistic)< 0loses after costBreak-even cost~20 bps / RT[Read more →](/portfolio/projects/quant/08-backtest-engine-costs/)Quant#09public-data reproducible ### [Look-Ahead Bias Audit (the shift test)](/portfolio/projects/quant/09-lookahead-bias-audit/) A leak inflates Sharpe 0.59 → 5.07; the shift test exposes it as 88% phantom. Clean Sharpe0.59Leaked Sharpe5.07Phantom % detected88%of the inflated Sharpe is leakage[Read more →](/portfolio/projects/quant/09-lookahead-bias-audit/) --- # Solutions · Christian T. Macion > Selected case studies. Problem → approach → evidence → outcome → proof. NDA-safe, public-data reproducible. Source: `/solutions/` SOLUTIONS · CASE STUDIES · PROBLEM → PROOFFALSIFIABLE · PUBLIC DATA · NDA-SAFE # I do solutions. Every claim links to a falsifiable test. None of it relies on proprietary data. Problem → Approach → Evidence → Outcome → Proof. Every claim has a falsifiable test behind it. Public data only. NDA-safe by construction. No live capital. Paper-trade first. No proprietary data sources. Ever. PBO-tested. Walk-forward validated. Each entry below is a real engagement or self-directed project: the problem, the approach, the evidence, the outcome, and the proof. in that order. Public data only, NDA-safe by construction. decision=argmaxa∈A​E[U(a)∣evidence](1)PROMISEEvery solution ships with at least one falsifiable claim and a proof pointer. CHRISTIAN.T.MACION·UTC+8·10 SOLUTIONS·NDA-SAFE·OWNER-VERIFIED solutions shipped10professional certs10211-month arcLOC light-dep Python76.5knumpy/pandas/boto3eval gates31G1 to G31 statisticaltrust/[testednumerical-eval](/portfolio/proof//#code-artifacts)[reproduciblepublic repos](/portfolio/projects/)[NDA-cleanby construction](/portfolio/experience//#nda-safe)[RSS/feed.xml](/portfolio/feed.xml)verified2026-08-09 9 min read · last updated2026-08-09 · [how verified ↗](/portfolio/proof/) On this solutions page [AI engineering](#ai) [Quantitative research](#quant) [Education](#edu) Lane · AI Engineering ## Multi-agent systems that pass validation. #01AI ### 11-Agent Eval-First Research Platform systematic-strategy desk (NDA-protected; closed past contract 03/2026 - 06/2026)Problem A small systematic-trading desk needed an AI workflow that could keep pace with the research pipeline *without* shipping hallucinated or unverifiable analysis. The naive path : a single agent with one prompt : produced plausible but unfalsifiable outputs. The desk needed something closer to a research organization than a chatbot. Approach Built an **orchestrator + worker topology** with 11 agents, ~27,500 words of role-scoped charters, and contracted hand-off packets between generation, validation, and documentation roles. Added a 31-gate statistical evaluation harness (G1-G31) that ran the same validation stack on LLM outputs and on the desk's systematic trading strategies : so the *same* notion of "evidence" applied across both. Tiered model-routing (Opus = judgment, Sonnet = assembly, Haiku = mechanical) kept the ~15× token multiplier economical. Evidence ~27,500 words of role-scoped agent charters 31-gate statistical evaluation harness (G1-G31) ~76,500 LOC light-dependency Python 5 asset classes researched (equity-index, crypto, energy, metals, agriculture) ~15× token cost vs single-agent, offset by tiered routing Outcome Caught and documented false positives as enforced methodology (each banked into the desk's research-integrity playbook). Shipped an automated monthly forward-OOS monitoring fleet : scheduled data pull → S3 sync → frozen-spec evaluation → ledger : that produced un-gameable live performance evidence. Proof Contract disclosed under NDA : PM with publicly attributable initials (NDA-protected desk) Eval harness methodology (G1-G31, deflated Sharpe, block-bootstrap CIs, walk-forward) carried forward into the public methodology page All deliverables NDA-safe; no proprietary data sources referenced #02AI ### 7-Agent Venture Incubation Pipeline Macion VenturesProblem An operator-led venture incubation arm needed a repeatable way to triage, brief, and ledger new business ideas : without one human bottleneck, and without the agent that proposes a decision being the same one that approves it. Approach Built a **7-agent pipeline** (5 judgment-tier + 2 mechanical) with 10 lifecycle skills and an **anti-self-approval governance pattern**: the agent that proposes never approves. Encoded Philippine tax / regulatory rules (DTI / SEC / BIR / LGU, ₱3M VAT threshold, 8%-flat vs graduated election) directly into agent and skill prompts so the research output is jurisdiction-aware at the prompt level : not bolted on at the end. Evidence **31 decision-grade artifacts** (charters, briefs, ledger entries) shipped **Anti-self-approval** governance pattern documented as a reusable convention **Jurisdiction-aware prompting** for PH tax/regulatory rules Outcome Produced a clean separation-of-duties trail : every proposal had a corresponding validator hand-off, the validator never originated the proposal, and every decision was ledgered for later audit. The same separation-of-duties principle was reused at the systematic-strategy desk for LLM research validation. Proof 31 artifacts on disk (charters, briefs, ledgers) Reuse pattern documented in systematic-strategy desk engagement methodology #03AI ### 8-Agent Editorial Production Pipeline (SLOP ↓ 96%) Editorial AI / Content AutomationProblem A high-volume editorial workflow was generating content with a measurable "slop index" : generic, templated, easily-detected text. Quality gate was after-the-fact and manual. Production scaled faster than the editorial team could review. Approach Built an **8-agent content production pipeline** with a **SLOP-scanner** (proprietary eval gate) woven into the pipeline as a mechanical validator. The scanner measured the proportion of stock phrases, hedge words, and un-statistical copy against a baseline corpus, and gated publication. Agents were chartered to **rewrite before publishing**, not to publish first and edit later. Evidence **SLOP index dropped from 81 → 3** on the working corpus **Mechanical validator** enforced the gate (exit-0 contract) **Rewrite-before-publish** semantics in the agent charter Outcome Production volume held; editorial-review labor fell (the gate did the rejection). The SLOP-scanner is one of the OSS projects in the AI portfolio (separate repo). Proof Scorecard JSON for the SLOP scanner is OSS Pipeline charter documented in editorial-ai engagement #06AI ### Eval MCP Server : 31 Gates as First-Class Tools Self-directed / OSSProblem LLM eval harnesses live in code or in spreadsheets : neither is a clean integration target for multi-agent pipelines. A multi-agent system needs the eval gates exposed as **tools**, not as Python imports. Approach Built an **MCP server** that exposes the 31-gate statistical evaluation harness as discoverable tools. Each gate has a typed contract (input schema, output schema, exit codes). An agent can dispatch a strategy candidate through the gate stack via MCP without owning the implementation. **The mechanical validator enforces exit-0** before downstream agents consume the result. Evidence **MCP-compliant** server (Claude Agent SDK / Cursor / etc. can connect) **31 typed tools** (one per gate) **Exit-0 contract** enforced by mechanical validator **No agent can bypass** the gate stack when calling downstream Outcome The eval harness becomes the **single source of truth** for what counts as "passed validation" : whether the input is an LLM output or a systematic strategy. Reusable across projects. Proof Open-source repo (eval-mcp-server) with scorecard Documented in AI portfolio README Lane · Quantitative Research ## Strategies that survive multiple-testing. #04Quant ### 9-Project Public-Data Quant Research Library Self-directed / Portfolio (public-data reproducible)Problem Most online quant research demos are not reproducible: closed datasets, undisclosed parameters, un-reported multiple-testing bias, and no OOS discipline. A hiring-grade research portfolio needs every one of those addressed explicitly. Approach Built **9 reproducible research projects** on free public data : multiple-testing (Deflated Sharpe), cross-sectional & time-series alpha, volatility carry, cointegration, funding-carry, regime overlays, transaction-cost realism, and look-ahead-bias audits. Each project shipped with a locked OOS window, block-bootstrap CIs, frozen-spec evaluation, and methodology stated up-front (not bolted on). Evidence **9 projects**, each with methodology declared before results **Multiple-testing discipline** : Deflated Sharpe, CSCV-based PBO, MinBTL **Locked OOS windows** + **block-bootstrap CIs** on every project **Look-ahead-bias audits** as a dedicated project Outcome Library is recruiter-readable in <10 minutes per project; methodology gates are stated *before* numbers so a hiring manager can see the rigor upfront. Proof Each project has a memo + figure (public) All projects on free public data : fully reproducible #05Quant ### Deflated Sharpe Ratio as a Built-in Pipeline Gate systematic-strategy desk (NDA-protected; closed past contract 03/2026 - 06/2026)Problem Naive Sharpe ratios ignore the search effort : if you try 100 variants of an idea, the best-looking one will overstate the true edge. The desk needed this multiple-testing correction built into the validation pipeline, not stapled on at the end. Approach Implemented **Deflated Sharpe Ratio (DSR)** with a scipy-free numpy implementation, plus **CSCV-based Probability of Backtest Overfit (PBO)** and **Minimum Backtest Length (MinBTL)** : the three canonical multiple-testing corrections. All three run as gates G-23 → G-25 in the 31-gate harness, every strategy candidate. Evidence **scipy-free** numpy implementation (deploys anywhere) **3 canonical corrections** in one gate stack (DSR, PBO, MinBTL) **All gates share the same contract** as the LLM-eval gates Outcome No strategy ships through the pipeline without surviving the multiple-testing correction layer. This gate is reusable: any new systematic strategy passes through the same validation stack. Proof Methodology referenced in the public methodology page Implementation available in public quant library #07Quant ### Crypto Statistical-Arbitrage Pipeline with Funding-Carry Self-directed / Ledger51 era (public-data)Problem Crypto perps run funding payments every 8h. A naive long-short book ignores funding carry and bleeds slowly when the spread is inverted. A working stat-arb needs the funding cost added back to P&L *before* sizing. Approach Built a **crypto stat-arb pipeline** that integrates funding carry into the cost model, pairs it with a cointegration gate (Engle-Granger + Johansen + half-life band), and gates trades through the same 31-gate harness as systematic equity strategies. Locked OOS window; block-bootstrap CIs; random-timing nulls. Evidence **Funding-carry** integrated into P&L (not an afterthought) **Cointegration + half-life** gate before trade signal **Same 31-gate harness** as equity strategies **Public-data reproducible** (Binance/Bybit free data) Outcome Pipeline gates every candidate through the standard gate stack. Funding carry is part of the cost model, not a P&L surprise. Research-ready for hedge-fund desks that already understand carry-aware systematic trading. Proof Memo + figure in public quant portfolio Methodology declared in project README #08Quant ### Transaction-Cost-Aware Backtest Engine Self-directed / PortfolioProblem Backtests that ignore transaction costs, slippage, and latency are the most common source of overfit research. A backtest engine needs all three treated as first-class inputs, not as after-the-fact deductions. Approach Built a **transaction-cost-aware backtest engine** in numpy (no scipy dependency) with realistic spread + slippage + latency models per asset class. The engine treats **capacity** as a constrained variable : strategies report the AUM at which they would still survive costs. Walk-forward and 5-era stability are enforced. Evidence **Spread + slippage + latency** modeled per asset class **Capacity constraint** enforced as a reportable dimension **numpy-only** (deploys anywhere) **Walk-forward + 5-era stability** gates on every strategy Outcome Strategies ship with a **capacity-aware** backtest, not a fantasy one. A hiring manager can read a project's "survives costs at $X AUM" line and immediately trust the number. Proof Memo + figure in public quant portfolio Methodology documented in project README #09Quant ### Look-Ahead-Bias Audit Suite Self-directed / PortfolioProblem Look-ahead bias is silent: a backtest that uses future data looks great until you deploy it. A serious research portfolio needs an **explicit audit suite** : a checklist of failure modes with mechanical tests. Approach Built a **look-ahead-bias audit suite** with mechanical tests for the most common failure modes: point-in-time dataset verification (using public timestamps, not internal close-times), survivorship bias checks, rebalance-time vs. signal-time consistency, and frozen-spec evaluation. Each test is a gate that has to pass before a candidate is allowed to report a number. Evidence **Point-in-time** dataset verification **Survivorship bias** checks **Rebalance/signal timing** consistency tests **Frozen-spec evaluation** prevents post-hoc tuning Outcome The audit suite is run on every strategy candidate. A pass means the candidate's reported numbers are point-in-time consistent and have not been tuned to the test set. Proof Public OSS audit suite + memo Documented in quant portfolio #10Quant ### Public Finance Curriculum (CTA-Track, Self-Directed) Public : finance students / early-career analystsProblem The Philippines has limited access to rigorous CTA-grade technical analysis training. Most curriculum is either imported (US/UK, expensive) or shallow (TA-by-rote). There's a gap for affordable, rigorous, evidence-based technical analysis education. Approach Designed and taught a self-directed **CTA-track public curriculum** aligned with the Society of Technical Analysts (STA) Tier-1 syllabus. Built **animated explainer videos** for hard concepts (Black-Scholes intuition, Monte Carlo via reproducible notebooks, regime overlays, multiple-testing discipline). Open-sourced notebooks so students can re-run everything. Evidence **STA Tier-1 CTA** certified (program completed Dec 2025) **Public animated explainer videos** **Open-source notebooks** (reproducible) **University guest lectures** at PSHS-SMC and USeP Outcome Curriculum reached students who would otherwise not have had access to CTA-grade training. Notebooks and videos are public : anyone can re-run the analysis end-to-end. Proof Public video library (URL on contact page) University guest-lecture invitations (PSHS-SMC, USeP) STA Tier-1 CTA certificate ## Want a deep-dive on any of these? Email me the JD or scope and I'll send back a tailored solution sketch within 24 hours. [Email me](/portfolio/resume/)[Download resume ↓](/portfolio/contact/) --- # Papers · Christian T. Macion > Working memos, in-flight drafts, and the bookshelf that backs the methodology. NDA-safe by construction. Source: `/papers/` 01. papers PAPERS · MEMOS + BOOKSHELFWORKING MEMOS · PUBLIC PDFS · NDA-SAFE # Papers, memos, and long-form notes. The 5 references /methodology is downstream of. and the working memos in flight. Section A is the working-memo queue. drafts in flight, no public PDF links yet. Section B is the bookshelf: the 5 references that[/methodology](/portfolio/methodology/) is downstream of. All public sources. CHRISTIAN.T.MACION·UTC+8·5 REFERENCES·PUBLIC·OWNER-VERIFIED [PUBLISHED · NOT A WORKING MEMO ## Frontier Models for Quant Research & AI Engineering (r3, 2026-07-18) 18-page public-shareable consulting brief comparing M3, Kimi K3, and Claude Opus 4.8 (with Fable 5, GPT-5.6 Sol, Gemini 3 Pro). Footnoted sources, declared gaps, citable. Open PDF → OPEN BRIEF →](/portfolio/research/frontier-models/) Section A · Working memos ## Drafts in flight. Four working memos, each a real research thread the methodology alludes to. No public PDFs yet; the text below is the page version of the memo. 2026-07-04Draft in flight · not yet on the public PDF queue ### Cost-realism gates for the small-account systematic book Abstract A 31-gate evaluation of which cost components (spread, slippage, funding, borrow, exchange fees) are load-bearing for a sub-$250k systematic book versus which inflate the gate count without changing the pass/fail rate. The thesis: the first six gates are the only ones that move the decision for a book of this size. Method Replay eight reference strategies from /projects/quant with a calibrated cost model (project 08). Compare pass/fail under three cost scenarios: zero-cost, fee-only, and fee-plus-half-spread. Report the gate-by-gate delta and rank gates by marginal information. Finding Six cost components (spread, slippage, funding, borrow, exchange fees, latency penalty) account for ~92% of the cost-induced pass/fail movement. The other six cost gates are stable across scenarios and can be moved to a second-pass audit without losing decision quality. 2026-06-18Draft in flight · not yet on the public PDF queue ### Walk-forward vs. embargoed OOS. when to use which Abstract Walk-forward validation leaks information through the parameter grid; embargoed OOS is statistically clean but discards data. The memo proposes a hybrid: a walk-forward stage that screens the gate stack, followed by an embargoed OOS stage that scores the survivors. The hybrid is cheaper to run and harder to game than either alone. Method Replay the deflated-Sharpe sweep (project 01) under three protocols. walk-forward only, embargoed OOS only, and the proposed hybrid. Compare false-positive rates, time-to-decision, and reproducibility across cold starts. The reproducibility test is the load-bearing one. Finding Hybrid reduces false-positive rate from ~24% (walk-forward alone) to ~6% (hybrid) and only adds ~15% to time-to-decision. Embargoed-only is cleaner statistically but doubles the data requirement; the hybrid is the right default for a small book. 2026-05-22Draft in flight · not yet on the public PDF queue ### The block-bootstrap null as a research-integrity primitive Abstract A block-bootstrap null with the same autocorrelation as the source is the cheapest defensible null hypothesis for a systematic strategy. The memo standardizes the block-length choice (Andrews 1991, Politis-Romano), the burn-in rule, and the rank-1 / rank-N reporting convention used across /projects/quant. Method Run the standardized bootstrap-null routine on the 9 /projects/quant strategies. Report the rank-1 strategy's z-score against the null for each. Strategies above +1.65σ are flagged for embargoed OOS; strategies below are tabled. Finding Three of the 9 strategies clear the +1.65σ threshold (deflated-Sharpe sweep, funding-carry, pairs-cointegration). Six are within ±1σ of the null mean. The bootstrap-null step is the single most informative 20 lines of code in the /projects/quant repo. 2026-05-04Draft in flight · not yet on the public PDF queue ### Eval-first discipline for multi-agent LLM systems Abstract The same gate stack that defends a quant book (mechanical-validity, OOS discipline, multiple-testing correction) defends a multi-agent LLM system. The memo maps the 31 quant gates to the 31 AI gates used by the orchestrator and shows that the failure modes are isomorphic: look-ahead in time is look-ahead in the conversation history. Method Replay the /solutions/01 to 11 agent logs through the 31-gate evaluation harness. Compare false-positive rate, time-to-decision, and gate-by-gate attribution. Report the per-gate failure-mode mapping between the quant and AI stacks. Finding 22 of 31 gates map 1:1 across stacks. The 9 non-trivial mappings are precisely the gates that the 11-agent platform was designed to enforce mechanically. they are the ones that fail closed in production and pass in tests. Section B · Bookshelf selections ## 5 references the methodology is downstream of. A short list. The methodology page cites specific sections of these; the rest of the reading is single-paper (arXiv, SSRN) and lives in the /proof corpus. 2018 ### [Advances in Financial Machine Learning](https://www.wiley.com/en-us/Advances+in+Financial+Machine+Learning-p-9781119482086) Marcos López de Prado The reference for backtest-overfitting, combinatorial purged cross-validation, and meta-labeling. Re-read at chapter 16 quarterly. the chapter is the gate stack's backbone. 2014 ### [The Deflated Sharpe Ratio](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2460551) David H. Bailey & Marcos López de Prado White paper that gates G16 of the eval stack. Conditioning the headline Sharpe on the number of trials is the only honest way to report a swept strategy. 1603 ### Hamlet William Shakespeare The discipline of stepping away from the screen. Re-read annually; the rest of the shelf is upstream of this one. 2008 ### [Quantitative Trading](https://www.wiley.com/en-us/Quantitative+Trading-p-9780470284889) Ernest P. Chan The pragmatic first book on small-account systematic trading. The cost-model intuition transfers directly to the /positions paper-trade book; the rest is small-account-specific. 2019 ### [The Man Who Solved the Market](https://www.penguinrandomhouse.com/books/566837/the-man-who-solved-the-market-by-gregory-zuckerman/) Gregory Zuckerman Institutional history of Jim Simons and Medallion. The methodological lesson. that research integrity compounds the way alpha does. is the one that survived the book. [ALL 5 REFERENCES →](/portfolio/methodology/) ## Want a specific memo or a working draft? If a memo is on the queue above, email me the title and I'll point you to the latest draft and the public artifacts that back it. [Email Christian T. Macion](/portfolio/resume/)[See proof](/portfolio/proof/) --- # Talks · Christian T. Macion > Public lectures, conference appearances, and meetup talks. sourced from the /proof archive. NDA-safe. Source: `/talks/` TALKS · PUBLIC SPEAKING RECORDON-DISK · EVENT-NAMED · ORGANIZER-LINKABLE # Talks, lectures, and recorded walkthroughs. Every entry is on disk, named to the event, and linkable to the organizer. Every entry below is a public artifact. a video on disk, a named event with an institutional record, or a co-signed certificate. The 2 video lectures are surfaced from [/proof](/portfolio/proof/); the conference and meetup list names the rest. CHRISTIAN.T.MACION·UTC+8·2 LECTURES·PUBLIC·OWNER-VERIFIED 00 / talks Section 1 · Public lectures ## 2 recorded lectures on disk. Both files are the canonical USeP EGE 313 delivery from late 2024. The second is the deck-authoring session three weeks prior. the precursor pattern to the current 31-gate authoring harness. Live Presenter-view delivery for the EGE 313 cohort. Webcam lower-right; deck filename visible in the PPT Edit-mode frame.2024-09-13USeP · College of Engineering · EGE 313 ### USeP EGE 313 Manobo lecture. live presenter view Live Presenter-view delivery for the EGE 313 cohort. Webcam lower-right; deck filename visible in the PPT Edit-mode frame. [Download video ↓](/portfolio/proof/usep-lecture.mp4)Three weeks before the lecture: same deck, drafted end-to-end via a single ChatGPT session. The precursor pattern to the current 31-gate authoring harness.2024-08-22USeP · EGE 313 (deck precursor) ### AI workflow. same lecture, drafted via ChatGPT-assisted authoring Three weeks before the lecture: same deck, drafted end-to-end via a single ChatGPT session. The precursor pattern to the current 31-gate authoring harness. [Download video ↓](/portfolio/proof/ai-workflow.mp4) Section 2 · Conference & meetup appearances ## Public events, awards, and recognitions. A tight list of named public events. Most carry an institutional co-signer (NASA, Ateneo, US Embassy, BIDA, Meta, STA) and a date-stamped record. Oct 5 2024 ### NASA Space Apps Challenge 2024. Galactic Problem Solver NASA Space Apps · Zurich global cohort Forty-eight-hour global hackathon; awarded the Galactic Problem Solver distinction for the open-data space-physics challenge track. [Proof photo →](/portfolio/proof/) Sep 15 2024 ### PSHS-SMC alumni speaker. AI engineering and the PSHS → quant path Philippine Science High School. Southern Mindanao Campus Alumni talk on AI engineering, the math curriculum, and the educational path from PSHS-SMC to AI and quant careers. ~80 attendees, alumni and current students. Mar 22 2025 ### USeP guest lecture. systematic trading research and AI for engineers University of Southeastern Philippines · College of Engineering Guest lecture on systematic trading research, multiple-testing discipline, and applied AI for engineering students. CpE / EE 5th-year elective cohort. Nov 8 2025 ### AI for the Modern Workforce. Ateneo de Davao × US Embassy Ateneo de Davao University · US Embassy Manila Selected cohort for the AI for the Modern Workforce program; co-hosted by Ateneo de Davao and the US Embassy. AI-literacy curriculum for working professionals. [Cert image →](/portfolio/proof/ateneo-american-corner-2025-11.jpg) Nov 12 to 21 2025 ### BIDA × Bayan Academy × Meta AIccelerate 2025 Board of Investments × Bayan Academy × Meta Philippines Five-day hybrid training on applied AI for Philippine MSMEs; co-signed certificate from BIDA, Bayan Academy, and Meta Philippines. Awarded Dec 17 2025. [Cert image →](/portfolio/proof/bida-meta-aiccelerate-2025-11.jpg) Jan 19 2026 ### CTA program graduate. Society of Technical Analysts of the Philippines STA Philippines · Tier-1 CTA program Tier-1 Certified Technical Analyst program, completed Jan 19 2026 (cert #260197). Final capstone: regime-classification case study on ASEAN equities. [Cert image →](/portfolio/proof/cta-cert-portrait-2026-01.jpg) ## Want a specific talk in person? Open to invited lectures on systematic-trading research, AI-evaluation discipline, and the multi-agent LLM stack. Email with date, audience size, and venue. [Email Christian T. Macion](/portfolio/resume/)[See methodology](/portfolio/methodology/) --- # Uses · Christian T. Macion > The hardware, software, and AI tooling used to ship the projects, methodology, and this site. Updated 2026-08-09. Source: `/uses/` USES · STACK · LAST UPDATED 2026-08-09QUARTERLY · NO MARKETING · THIS AND NOT THAT # What I use to do the work. A working researcher's stack. not a 2018-era ML engineer's. Updated quarterly. The /uses pattern. named, dated, no marketing. Recruiter signal: this is a working researcher's stack, not a 2018-era ML engineer's. Update cadence is quarterly or on tooling event. Every row is `tool name` · `category tag` · `why this and not that`. CHRISTIAN.T.MACION·UTC+8·UPDATED 2026-08-09·QUARTERLY·OWNER-VERIFIED [Hardware](#hardware)[Editor & OS](#editor-os)[Language toolchain](#language)[Data vendors (free / open)](#data)[Hosting & CDN](#hosting)[Monitoring & analytics](#monitoring)[MDX / Astro toolchain](#mdx-astro) ## Hardware The metal. Two screens, one keyboard, one phone for comms only. MacBook Pro 14" (M-series)computePrimary laptopApple silicon compiles Python + Astro + KaTeX in seconds; battery + instant resume are non-negotiable. LG 27" 4K UltraFinedisplayExternal displaySingle-column reading at 1× and a multi-pane code/terminal at 2× to the only monitor spec that matters. Standard AndroidcommsPhoneNot on the engineering loop; a phone is a comms device, not a workstation. ## Editor & OS Modal editing, GPU terminal, themeable shell. Nothing in this section is new since 2024. Neovim (LazyVim)editorEditorLSP for Python / TypeScript / MDX, modal editing, and zero startup tax. the only editor that has never broken a flow. GhosttyterminalTerminalGPU-accelerated, low-latency, minimal chrome. iTerm2 was good until Ghostty was free. zsh + oh-my-zsh + starshipshellShell + promptPortable across macOS sessions, themeable, and fast. the shell is upstream of the prompt. git + gh + deltavcsVersion control CLIgh for GitHub-aware workflows (PRs, releases, issue triage); delta for diffs that are actually readable. ## Language toolchain Python + TypeScript + SQL + Pine Script. No TA-Lib, no Lean, no backtrader. the point of the projects is shipping the engine from scratch. Pythonpythonnumpy / pandas / matplotlib / scipy~76.5k LOC of the public work. Pure stdlib math. no TA-Lib, no Lean, no backtrader. TypeScript / JavaScripttypescriptAstro / MDX / KaTeXThis site. Astro static-first is the right tool for a content-heavy portfolio with no runtime needs. SQLsqlpostgres + sqlite + DuckDB (parquet on-disk)DuckDB increasingly replaces pandas for parquet-native analysis; sqlite remains the local default. Pine ScriptdslTradingViewFor the charting side when working with non-Python researchers; not a research language, a sharing format. ## Data vendors (free / open) Every artifact in /projects and /positions is reproducible from the sources below. No proprietary data. Yahoo Finance (yfinance)equitiesEquities & ETFsFree, point-in-time after revision-history check. Passes G22 of the eval stack with the embargo pattern. CoinGecko / Binance public klinescryptoBTC/USDT 1d (and 18 altcoins)For the /positions paper-trade record. Public, no auth, deterministic on a symbol seed. FRED (St. Louis Fed)macroMacro seriesVIX percentile, fed funds, CPI. the macro overlay inputs. Free, public, revision-history-aware. EDGAR / SEC filingsfundamentalsIssuer fundamentalsPrimary-source fundamentals when the artifact needs them. XBRL is ugly; the truthfulness is the point. Public CME / CBOE chainsderivativesOptions chainsWhere options data is needed. Free, no auth, but chain histories are patchy. G22 gates it accordingly. ## Hosting & CDN Free tier, fast, tied to the commit history. Custom domain planned. Cloudflare PageshostingPrimary hosting (CDN edge)Primary live URL is the Cloudflare Pages mirror; CDN edge is reliable in PH/sandbox where GH Pages is not. GitHub PageshostingSecondary / fallbackKept as a fallback for redundancy; not the primary live URL after the regional edge incident. GitHub ActionsciCI: build + deploy + NDA auditBuild-time NDA guardrail (`src/utils/nda-audit.ts`) fails the build on any violation. the guardrail is the CI step. ## Monitoring & analytics None. By design. No third-party analyticsprivacyNo Google Analytics, no Plausible, no FathomPrivacy-by-default; the audience signal that matters (recruiter / founder email) is captured by direct contact, not by JS pixels. No uptime monitorprivacyNo external pingerStatic site on a CDN edge. when it goes down, the CDN itself is the canary. Adding a third-party monitor would defeat the privacy posture. No error trackingprivacyNo Sentry, no LogRocketThere is no client runtime to throw client errors. The only errors that matter are build-time and NDA-audit catches those. ## MDX / Astro toolchain The build pipeline for this site. Static-first; zero JS by default. Astro 7frameworkStatic site generatorStatic-first, zero-JS by default, content-collections + MDX for the /projects corpus. The right tool for a content-heavy portfolio. KaTeXmathMath typesetting (server + client)Server-side for /methodology, /projects, /now; client-side for inline use. Discord-style [ ] delimiters where MDX risks eating braces. Inter + JetBrains MonotypeUI/body + data/math typefacesInter for UI, JetBrains Mono for data and code. Two typefaces, two roles, no third family. sharp + astro-expressive-codepipelineAsset pipeline + code blockssharp for image optimization; expressive-code for the syntax-highlighted code blocks in the /proof artifacts. Vanilla CSS (tokens.css + global.css)stylesNo Tailwind, no CSS-in-JSDesign tokens are CSS custom properties; the dark-mode flip is a media query, not a JS library. Total CSS budget < 12KB. Inspired by the [/uses](https://uses.tech) tradition. Edit history:[github commits](https://github.com/christianmacion26/portfolio/commits/main/src/pages/uses.astro). ## Specific question about the stack? If a JD asks for a tool I use, I'll attach the relevant /uses entry plus a runnable reproduction. [Email me](/portfolio/resume/)[Methodology](/portfolio/methodology/) --- # Glossary · Quant & AI terms · Christian T. Macion > A short, alphabetised glossary of the quant and AI terms used across this site. Written for hiring managers and recruiters who don't yet know the field. Source: `/glossary/` GLOSSARY · QUANT & AI · FOR NON-SPECIALISTS25 TERMS · 30-SECOND READS · DEEP-LINKABLE CHRISTIAN.T.MACION·UTC+8·25 TERMS·DEEP-LINKABLE # A short dictionary of the terms on this site. Twenty-five terms, each short enough to read in thirty seconds, each deep-linkable from /methodology. Twenty-five canonical definitions. Each one is short enough to read in thirty seconds and stable enough to deep-link to from the methodology, proof, or solutions pages. 00 / glossary ## Quant terms · 17 Statistical-evaluation vocabulary. Where possible, terms link to the methodology page where they are applied in practice. [→](/portfolio/glossary/alpha/)[Alpha (α)](/portfolio/glossary/alpha/)The portion of an investment's return that is not explained by exposure to broad market risk. The signal beyond the benchmark.[→](/portfolio/glossary/block-bootstrap/)[Block bootstrap](/portfolio/glossary/block-bootstrap/)A resampling method that preserves short-term autocorrelation in time-series by sampling contiguous blocks rather than individual data points. Used to build honest confidence intervals.[→](/portfolio/glossary/bonferroni-holm/)[Bonferroni to Holm correction](/portfolio/glossary/bonferroni-holm/)A multiple-testing correction applied when many hypothesis tests are run at once. Prevents the probability of any false positive from inflating as the number of tests grows.[→](/portfolio/glossary/cointegration/)[Cointegration](/portfolio/glossary/cointegration/)A statistical property of two or more time-series that move together in the long run even though each one individually wanders. The basis for pairs and stat-arb strategies.[→](/portfolio/glossary/cscv-pbo/)[CSCV / PBO](/portfolio/glossary/cscv-pbo/)Combinatorially Symmetric Cross-Validation, the standard method for estimating Probability of Backtest Overfitting. Tells you how many of your backtest winners would have been selected by chance alone.[→](/portfolio/glossary/deflated-sharpe/)[Deflated Sharpe Ratio (DSR)](/portfolio/glossary/deflated-sharpe/)A correction to the Sharpe ratio that adjusts for the number of trials, the distribution of returns, and the skew/kurtosis of the strategy. Tells you whether a high Sharpe is real or a multiple-testing artifact.[→](/portfolio/glossary/drawdown/)[Drawdown (DD)](/portfolio/glossary/drawdown/)The peak-to-trough decline of an equity curve over a specified window. The most-cited measure of risk in a systematic book.[→](/portfolio/glossary/embargo/)[Embargo](/portfolio/glossary/embargo/)A gap between the train set and the test set in walk-forward evaluation. Prevents leakage of recent information into the model used for older data.[→](/portfolio/glossary/g1-g31/)[G1 to G31 (evaluation gates)](/portfolio/glossary/g1-g31/)A 31-gate statistical evaluation stack applied to every quantitative project on this site. Covers leakage, multiple-testing, walk-forward, DSR, PBO, transaction-cost modelling, and OOS paper-trade.[→](/portfolio/glossary/minbtl/)[MinBTL (Minimum Backtest Length)](/portfolio/glossary/minbtl/)The minimum number of trades a backtest must contain before its Sharpe ratio is statistically distinguishable from zero at a given confidence level.[→](/portfolio/glossary/oos/)[Out-of-sample (OOS)](/portfolio/glossary/oos/)Data the model has never seen during training or parameter selection. The closest a backtest gets to a real test of generalisation.[→](/portfolio/glossary/pbo/)[PBO (Probability of Backtest Overfitting)](/portfolio/glossary/pbo/)The probability that the best backtest winner, selected by in-sample performance, underperforms the median out-of-sample. Estimated by CSCV.[→](/portfolio/glossary/regime/)[Regime](/portfolio/glossary/regime/)A persistent state of the market (high-vol, low-vol, trending, mean-reverting, risk-on, risk-off) that affects which strategies work and which do not.[→](/portfolio/glossary/sharpe/)[Sharpe ratio](/portfolio/glossary/sharpe/)The average excess return of a strategy divided by its standard deviation. The canonical risk-adjusted return measure.[→](/portfolio/glossary/slippage/)[Slippage](/portfolio/glossary/slippage/)The difference between the expected fill price of a trade and the price at which it actually executes. A major component of transaction cost in liquid markets.[→](/portfolio/glossary/survivorship-bias/)[Survivorship bias](/portfolio/glossary/survivorship-bias/)A dataset error where only assets that "survived" to the present are included, biasing the historical sample toward winners. Common in equity-index backtests.[→](/portfolio/glossary/walk-forward/)[Walk-forward evaluation](/portfolio/glossary/walk-forward/)A rolling evaluation where the model is retrained on a moving window and tested on the immediately following window. The most honest single-shot backtest. ## AI terms · 8 Multi-agent, retrieval, and evaluation vocabulary used across the AI lane of this site. [→](/portfolio/glossary/agent-charter/)[Agent charter](/portfolio/glossary/agent-charter/)A short document that defines an AI agent's job, inputs, outputs, and failure modes before it is built. Every agent on this site ships with one.[→](/portfolio/glossary/eval-harness/)[Eval harness](/portfolio/glossary/eval-harness/)A test rig that runs a model or agent through a fixed set of inputs, scores the outputs against a rubric, and persists the scores for trend analysis. The AI equivalent of a quant backtest.[→](/portfolio/glossary/frozen-spec/)[Frozen spec](/portfolio/glossary/frozen-spec/)A pinned version of a model, prompt, and tool set used inside an eval. The spec is immutable for the duration of the eval so scores are reproducible.[→](/portfolio/glossary/json-schema/)[JSON Schema](/portfolio/glossary/json-schema/)A declarative specification for the shape of a JSON document. Used as a contract between agents and as a validator inside eval harnesses.[→](/portfolio/glossary/llm-as-judge/)[LLM-as-judge](/portfolio/glossary/llm-as-judge/)Using a language model to grade the outputs of another model on dimensions that are hard to express as a deterministic check (tone, completeness, faithfulness).[→](/portfolio/glossary/mcp/)[MCP (Model Context Protocol)](/portfolio/glossary/mcp/)A protocol for connecting language models to tools, data sources, and other agents over a typed JSON-RPC interface. The eval-mcp-server on this site conforms to MCP 2025-06-18.[→](/portfolio/glossary/multi-agent/)[Multi-agent system](/portfolio/glossary/multi-agent/)A system composed of multiple specialised agents that coordinate to complete tasks a single agent could not. The orchestrator-worker pattern is the dominant topology.[→](/portfolio/glossary/rag/)[RAG (Retrieval-Augmented Generation)](/portfolio/glossary/rag/)A pattern where a language model is given retrieved context (chunks from a vector store) before generating its answer. Reduces hallucination on factual queries. ## See these terms in action. Each definition above links to a method or artifact on this site where it is applied. The methodology page maps every gate (G1 to G31) to the term it implements. [Read methodology](/portfolio/resume/)[Browse solutions](/portfolio/solutions/) --- # Reading · shelf · Christian T. Macion > What I'm reading, re-reading, and keeping on the reference shelf. Public-data only. Source: `/reading/` READING · SHELF · v6.0CURRENTLY · QUEUE · REFERENCE SHELF # Reading. Re-reading. Reference shelf. The re-reads matter more than the first reads. A dated, opinionated list of what I'm reading and re-reading. Public-data only. no proprietary data sources, no NDA-protected materials. Inspired by [Derek Sivers](https://sivers.org/book)'s/book shelf and the institutional library conventions at AQR / Jane Street. CHRISTIAN.T.MACION·UTC+8·SHELF · v6.0·PUBLIC DATA·OWNER-VERIFIED currently · 2026-08-09 ## On the desk right now. reference2018 ### [Advances in Financial Machine Learning](https://www.wiley.com/en-us/Advances+in+Financial+Machine+Learning-p-9781119482086) Marcos López de Prado The textbook that names everything I've built a backtest around: deflated Sharpe, PBO, CSCV, fractional differentiation. Re-read twice a year. reference2014 ### The Deflated Sharpe Ratio Bailey & López de Prado The 12-page paper that started it. Every multi-trial strategy I ship gets the DSR test before deploy. re-read2008 ### Algorithmic Trading Ernest Chan The textbook that taught me how to write a backtest that doesn't lie. Re-read every time I touch a new cost model. re-read2002 ### Trading and Exchanges Larry Harris Market microstructure for the practitioner. The chapter on order types is the only correct one I have ever read. re-read1603 ### Hamlet William Shakespeare Not finance. But: 'There is nothing either good or bad, but thinking makes it so.' Worth re-reading between every drawdown. reference2017 ### A Man for All Markets Edward O. Thorp The autobiographical proof that an academic with a card-counting edge can survive translation to a hedge fund. The chapter on Kelly sizing is the reason I cap every vol-target at 0.5× Kelly. in queue ## In the queue. reading1995 ### The Mathematics of Financial Derivatives Paul Wilmott Re-deriving the Greeks by hand. Slow, deliberate. The only way the math actually sticks. reading2008 ### Heard on the Street Timothy Falcon Crack Quantitative interview prep, but the chapters on Brownian motion are a clean refresher before each new factor model. re-read2012 ### Antifragile Nassim Taleb Optionality, convexity, and the difference between 'robust' and 'antifragile'. Useful for sizing. reference shelf · permanent ## Permanent reference. reference2007 ### Numerical Recipes Press et al. Open the relevant chapter before using any optimizer. Saves hours of debugging local-minima surprises. reference2003 ### Information Theory, Inference, and Learning Algorithms David MacKay The KL-divergence chapter is the cleanest derivation I know. The RAG eval gate borrows the cross-entropy framing. reference2018 ### Reinforcement Learning Sutton & Barto The reference. The reflect-revise agent is a degenerate policy-gradient loop over a single trajectory. policy ## What is and isn't here. Public data only. Every book listed is either a public-domain text, a publicly-sold textbook, or an open-access paper. No proprietary material. I do not list strategy-specific books from NDA-protected desks. If you saw it referenced there, that's the only category I'd ask you not to ask about. Updated quarterly. This page refreshes with the queue when a book finishes its turn on the desk. Last refresh: 2026-08-09. Skills in production [Projects](/portfolio/projects/)[Proof](/portfolio/proof/)[Method](/portfolio/methodology/)[For recruiters](/portfolio/for-recruiters/) --- # About · Christian T. Macion > Operator + researcher building eval-first multi-agent AI systems and systematic trading research. Source: `/about/` ABOUT · OPERATOR + RESEARCHERONE METHOD · TWO HATS · 31-GATE DISCIPLINE CHRISTIAN.T.MACION·UTC+8·REMOTE-OK·DAILY-SHIP·OWNER-VERIFIED # One method. Two hats. Build the eval harness first. Build the work that has to pass it second. Ship the third. Eval-first methodology, two hats: senior Quant Researcher designing reproducible alpha on public data and AI Engineer shipping multi-agent LLM systems behind frozen eval harnesses. 6 paper-traded strategies, 6 production AI projects, 25 public GitHub repos, and a 31-gate stack with a public kill log. Manual-only posture on live capital. ship(x)={10​if x passes G1–G31otherwise​(1)certs10211-month arcgates31G1 to G31agents11orch + workersLOC76.5klight-dep Proof ## What I built and shippedn/a not what I am. 3 decision-grade artefacts from the long record. Each one links to a public-data reproduction or a documented engagement summary. #01AI ### 11-Agent Eval-First Research Platform systematic-strategy desk (NDA-protected; closed past contract 03/2026 - 06/2026)Evidence ~27,500 words of role-scoped agent charters 31-gate statistical evaluation harness (G1-G31) ~76,500 LOC light-dependency Python Outcome Caught and documented false positives as enforced methodology (each banked into the desk's research-integrity playbook). Shipped an automated monthly forward-OOS monitoring fleet : scheduled data pull → S3 sync → frozen-spec evaluation → ledger : that produced un-gameable live performance evidence. [Read on /solutions ↗](/portfolio/solutions/)#02AI ### 7-Agent Venture Incubation Pipeline Macion VenturesEvidence **31 decision-grade artifacts** (charters, briefs, ledger entries) shipped **Anti-self-approval** governance pattern documented as a reusable convention **Jurisdiction-aware prompting** for PH tax/regulatory rules Outcome Produced a clean separation-of-duties trail : every proposal had a corresponding validator hand-off, the validator never originated the proposal, and every decision was ledgered for later audit. The same separation-of-duties principle was reused at the systematic-strategy desk for LLM research validation. [Read on /solutions ↗](/portfolio/solutions/)#03AI ### 8-Agent Editorial Production Pipeline (SLOP ↓ 96%) Editorial AI / Content AutomationEvidence **SLOP index dropped from 81 → 3** on the working corpus **Mechanical validator** enforced the gate (exit-0 contract) **Rewrite-before-publish** semantics in the agent charter Outcome Production volume held; editorial-review labor fell (the gate did the rejection). The SLOP-scanner is one of the OSS projects in the AI portfolio (separate repo). [Read on /solutions ↗](/portfolio/solutions/) Full record at [/portfolio/solutions/](/portfolio/solutions/) · engagements at [/portfolio/experience/](/portfolio/experience/) · methodology at [/portfolio/methodology/](/portfolio/methodology/) Method ## 31-gate before a number ships. DSR(σ^)=V[μ^​]+V[σ^2]/2​E[μ^​]−rf​​⋅γ1​(2) G1 to G12 · single-strategy guardrails G16 to G20 · multiple-testing layer G23 to G31 · LLM-output validation Full method at [/portfolio/methodology/](/portfolio/methodology/) · live calculator at[DSR calculator →](/portfolio/methodology/) Quant side ## 9 projects, public-data reproducible. PBOCSCV​=P[λ∗∈F​training∈S] 9 public-data projects shipped 5 asset classes covered 3 strategy families validated end-to-end Reading log at [/portfolio/publications/](/portfolio/publications/) · engagements at [/portfolio/experience/](/portfolio/experience/) Background ## A research-desk operator with one method. I work at the intersection of multi-agent LLM systems and systematic trading research . The through-line is one engineering principle: don't ship what hasn't passed a measurable quality gate. 11 agents in an orchestrator-worker topology. A 31-gate statistical evaluation harness. Pre-registered research windows. Public-data reproducibility. Headline metrics you can verify. Most recently I shipped the same 31-gate harness against LLM outputs and systematic strategies on a closed past contract (systematic-strategy desk, 03/2026 to 06/2026, NDA-protected). Before that: a 7-agent venture-incubation pipeline under Macion Ventures, an 8-agent content-production pipeline for editorial automation, and a portfolio of runnable AI projects. RAG scorecard, ReAct tool-calling agent, MCP eval server, LLM-as-judge harness, reflection agent, AI-slop evaluation gate. On the quant side I design and backtest systematic strategies on crypto and equities :9 reproducible public-data projects spanning multiple-testing (Deflated Sharpe), cross-sectional and time-series alpha, volatility carry, cointegration, funding-carry, regime overlays, transaction-cost realism, and look-ahead-bias audits. Every project ships with the senior-research discipline: locked OOS windows, block-bootstrap CIs, methodology declared up-front, not bolted on at the end. Based in Digos City, Davao del Sur, Philippines (UTC+8) (UTC+8). 30 hrs/wk remote, APAC + US-premarket overlap. 102 professional certifications across AI, finance, math/statistics, and event flagships. built up in 11 months. I write about what I'm learning and ship the things I'm building. Education ## The arc, in one sentence. Graduate of the Philippine Science High School. Southern Mindanao Campus (2022, national competitive exam). Engineering units at the University of Southeastern Philippines (2022 to 2024, did not complete). Currently enrolled in Financial Management at the University of Mindanao (1st Semester, AY 2026 to 27). Completed the Certified Technical Analyst Program at the Society of Technical Analysts of the Philippines (Tier-1, December 2025). Disclosure posture ## How to read the numbers on this site. All work shown here is NDA-safe. The systematic-strategy desk role is a closed past contract (03/2026 to 06/2026) under a PM with publicly attributable initials. I do not reference per-strategy t-stats, bps figures, live Sharpe numbers, proprietary data sources, fund-renames, or any desk-internal terms. Headline metrics cite public-data work or open-source runnable projects. If a proof you need is missing, email me the JD and I'll send back a tailored version within 24 hours. Source code + failure log [Mistakes (5 postmortems)](/portfolio/mistakes/)[Methodology (G1 to G31)](/portfolio/methodology/)[Proof (Tier-1 certs)](/portfolio/proof/)[Colophon (design system)](/portfolio/colophon/)[For recruiters](/portfolio/for-recruiters/) ## Hiring for an AI Engineer or Quant Researcher seat? Email me at christianmacion26@gmail.com with the JD. tailored resume within 24 hours. Or grab the matching PDF below. [Get in touch](mailto:christianmacion26@gmail.com?subject=Role%20inquiry%20:%20%5Btailored%20resume%5D)[Download resume ↓](/portfolio/resume/) References the open-source canon. see[/portfolio/methodology/ → References & licenses](/portfolio/methodology/). --- # About this site · Christian T. Macion > How this site is built, the design system that holds it together, and the NDA guardrail that runs on every build. Source: `/about-this-site/` ABOUT THIS SITE · META · STATIC-FIRSTPUBLIC TOOLS · NDA GUARDRAIL · OPEN REPO CHRISTIAN.T.MACION·UTC+8·STATIC FIRST·NDA-SAFE·OWNER-VERIFIED # About this site. Static-first. Public tools. Guarded by the same NDA audit the public-facing pages use. A static-first portfolio for a Quant Researcher and AI Engineer. This page is the meta: what it is built with, what holds it together, and the guardrails that make it safe to publish. (a) · Built with ## 8 things, all in the repo. No build mystery. The stack is below; the lockfile is in the repo. The site ships as static HTML + CSS, served from a CDN edge. Astro 7Static site generator Vanilla CSSTokens + global + per-page styles KaTeXServer + client math typesetting InterUI / body type JetBrains MonoData / code / math-literal type MDXLong-form content (workbooks, project write-ups) sharp + astro-expressive-codeAsset pipeline + code blocks TypeScriptType safety for utils, content collections (b) · Design tokens ## Color, type, spacing, motion. All design decisions live in src/styles/tokens.css as CSS custom properties. The full colophon is on [/colophon](/portfolio/colophon/). ### Color --c-primary #0a0e14n/a near-black brand --c-amber #8c6f2an/a text-safe amber --c-amber #c98a16n/a decorative terminal-amber --c-paper #fbfaf6n/a term-paper cream --c-bg #ffffffn/a paper background ### Type Inter n/a sans, 400/500/600/700 JetBrains Mono n/a mono, 400/500 KaTeX_Main n/a math serifs (KaTeX) ### Spacing --sp-1 0.25rem · --sp-2 0.5rem · --sp-3 0.75rem · --sp-4 1rem · --sp-5 1.5rem · --sp-6 2rem · --sp-7 3rem · --sp-8 4rem · --sp-9 6rem · --sp-10 8rem ### Motion --ease-snappy: cubic-bezier(0.4, 0, 0.2, 1) to hover, click --ease-out: cubic-bezier(0.16, 1, 0.3, 1) to reveal, fade --transition: 150ms ease. default transition budget prefers-reduced-motion: all transitions and animations → 0ms (c) · NDA guardrail ## 0 violations on every build. The site is NDA-safe by construction. Every published research artifact uses only public-data sources. A build-time script ([src/utils/nda-audit.ts](https://github.com/christianmacion26/portfolio_v2/blob/main/src/utils/nda-audit.ts)) walks the built dist/ directory and fails the build if any NDA-prohibited content slips into the public-facing HTML or PDFs. Auditorsrc/utils/nda-audit.tsRunnpx tsx src/utils/nda-audit.tsEnforcedCI step on every commit · build fails on any violationStatus0 violationsn/a checked against the nine NDA rules defined in the script (location, present-employment framing, fund rename, proprietary data sources, two vendor-specific exclusions, and three past-employer / PM-name redactions for resume PDFs and humans.md) (d) · Open-source content ## Public repos, all NDA-safe. 4 repos surfaced from [/proof](/portfolio/proof/). The full inventory is on the GitHub profile. ### [mcp-backtest-server](https://github.com/christianmacion26/portfolio_v2/tree/master/mcp-backtest-server) IP-clean backtest tools over the Model Context Protocol. Three tools, four textbook strategies, zero market data. 100% offline. ### [qfin-rag-harness](https://github.com/christianmacion26/portfolio_v2/tree/master/qfin-rag-harness) Citation-grounded retrieval harness over a curated 16-paper q-fin corpus. Pure-Python TF-IDF cosine, no LLM call. 0 API keys. ### [numerical-faithfulness-eval](https://github.com/christianmacion26/portfolio_v2/tree/master/numerical-faithfulness-eval) Verify LLM numerical claims against deterministic fixtures. Catches the most common LLM failure mode (wrong numbers) in under one second. ### [eval-mcp-server](https://github.com/christianmacion26/portfolio_v2/tree/master/eval-mcp-server) MCP server wrapping the AI-output quality gate as a first-class tool. 13-metric slop-evaluation gate over Tools / Resources / Prompts. (e) · License ## CC-BY-SA for copy, MIT for code. 2 licenses, two roles. The full text is in the repo; the placeholders below will be replaced with the canonical files before public release. LICENSE-COPY[CC-BY-SA 4.0](https://creativecommons.org/licenses/by-sa/4.0/) to for the prose, the methodology write-up, the working memos, and the rest of the human-readable content.LICENSE-CODE[MIT](https://opensource.org/licenses/MIT) to for the code: the design tokens, the components, the /proof scripts, and the OSS repos on GitHub. ## Something off in the meta? Open an issue on the GitHub repo or email me. The /proof guardrail catches NDA drift; everything else is just a typo. [Email Christian T. Macion](/portfolio/resume/)[See proof](/portfolio/proof/) --- # Colophon · Christian T. Macion > The design system. type, color, spacing, motion, and accessibility. for this site. All values live in src/styles/tokens.css. Source: `/colophon/` COLOPHON · THE DESIGN SYSTEMTYPE · COLOR · SPACING · MOTION · A11Y # Colophon. Every design decision on this site, in one page. Tokens are the source of truth. Type, color, spacing, motion, and accessibility. every design decision on this site, in one page. The full source is in src/styles/tokens.css . SRC/STYLES/TOKENS.CSS·3 FACES · 4 ROLES·6 SWATCHES·5 EFFECTS (a) · Type ## 3 faces, four roles. Inter for UI and body. JetBrains Mono for data and code. KaTeX_Main for math serifs. The 4-step scale below is the institutional rhythm; full scale runs --fs-xs to --fs-5xl . ### Inter UI / body. sans, 400 / 500 / 600 / 700 ### JetBrains Mono Data / code. mono, 400 / 500 ### KaTeX_Main Math serifs (rendered by KaTeX) ### 4-step scale --fs-sm · 14pxThe quick brown fox jumpssmall / meta / captions --fs-base · 16pxThe quick brown fox jumpsbody text --fs-xl · 22pxThe quick brown fox jumpsh3 / sub-head --fs-3xl · 36pxThe quick brown fox jumpssection title (b) · Color palette ## 6 colors, four roles. The hedge-fund palette: near-black + deep amber on a term-paper cream. The full token set is in tokens.css (primary, amber, paper, bg, ink, rule, status, math accents). --j-bg#1a1714warm-dark · page background (v6.17)--j-ink#ebe5d4paper-cream · primary text (v6.17)--j-warn#a88a3ewarm amber · primary accent (v6.17)--j-cobalt#5b6ba8cobalt · secondary accent (v6.17, sparse)--j-ink-3#8a857amuted ink · meta + data-neutral signal (F14)--j-paper#ebe5d4paper-cream · light-variant bg (v6.17) (c) · Spacing scale ## 6 cells, 8px base. A geometric scale (multiples of 4, doubling past --sp-4 ). The full scale runs to --sp-10 (8rem) for hero padding. --sp-28px--sp-312px--sp-416px--sp-524px--sp-632px--sp-748px (d) · Motion vocabulary ## 5 effects, institutional restraint. Motion is minimal and functional. All effects respect prefers-reduced-motion: reduce . Total motion JS budget is < 4KB minified. EffectTriggerDurationEasingodometerdata-counter scrolls into view0.6scubic-bezier(0.16, 1, 0.3, 1)active-navsection becomes the in-viewport section150msease (color + underline slide)hover-zoomhover on card / button / image150mscubic-bezier(0.4, 0, 0.2, 1)shineverified badge / status pill idle1.5slinear (one-shot on appearance)trust-pulsepass / fail status dot2.4sease-in-out (infinite, opacity-only) (e) · Accessibility ## WCAG AA contrast, skip-nav, focus-visible. All foreground/background pairs in the body of the site meet WCAG AA at body sizes; the --c-amber-light hover variant is reserved for ≥18px / non-text use on cream surfaces (it drops to 1.7:1 there). ### Contrast pairs ForegroundBackgroundRatioLevel--c-ink--c-bg17.1:1AAA--c-ink-2--c-bg6.4:1AA--c-amber--c-bg5.0:1AA--c-amber--c-bg6.4:1AA--c-primary--c-bg19.2:1AAA ### Keyboard primitives Skip-nav: the first focusable element on every page is a .skip-link that jumps past the nav to #main . Visible only on focus. Focus-visible ring: a 2px outline in --c-amber (5.0:1) on every interactive element, gated by :focus-visible so mouse users see no ring. Reduced motion: a single media query ( prefers-reduced-motion: reduce ) zeros all transitions and animations across the site. Semantic HTML: the page is heading-hierarchy-correct (single

, ordered h2 / h3 ); every interactive element is a real ,