https://christianmacion.github.io/portfolio/ Newest projects, research artifacts, and client solutions. Quant, AI, OSS. en Sun, 09 Aug 2026 03:08:15 GMT christianmacion26@gmail.com (Christian T. Macion) https://christianmacion.github.io/portfolio/solutions/#11-agent-eval-platform tag:christianmacion26.github.io,2026:solution:11-agent-eval-platform Sun, 09 Aug 2026 03:08:15 GMT 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. multi-agent eval-harness MCP model-routing walk-forward deflated-sharpe https://christianmacion.github.io/portfolio/solutions/#7-agent-venture-pipeline tag:christianmacion26.github.io,2026:solution:7-agent-venture-pipeline Sun, 09 Aug 2026 03:08:15 GMT 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. multi-agent governance venture prompts separation-of-duties https://christianmacion.github.io/portfolio/solutions/#8-agent-editorial-pipeline tag:christianmacion26.github.io,2026:solution:8-agent-editorial-pipeline Sun, 09 Aug 2026 03:08:15 GMT 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. multi-agent content eval-gate slop-scanner https://christianmacion.github.io/portfolio/solutions/#9-project-quant-library tag:christianmacion26.github.io,2026:solution:9-project-quant-library Sun, 09 Aug 2026 03:08:15 GMT 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. deflated-sharpe PBO CSCV walk-forward block-bootstrap multiple-testing https://christianmacion.github.io/portfolio/solutions/#crypto-stat-arb-funding tag:christianmacion26.github.io,2026:solution:crypto-stat-arb-funding Sun, 09 Aug 2026 03:08:15 GMT 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. crypto stat-arb funding-carry cointegration half-life https://christianmacion.github.io/portfolio/solutions/#cta-curriculum-public tag:christianmacion26.github.io,2026:solution:cta-curriculum-public Sun, 09 Aug 2026 03:08:15 GMT 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. education STA CTA public-curriculum open-source https://christianmacion.github.io/portfolio/solutions/#deflated-sharpe-gate tag:christianmacion26.github.io,2026:solution:deflated-sharpe-gate Sun, 09 Aug 2026 03:08:15 GMT 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. deflated-sharpe PBO MinBTL multiple-testing scipy-free https://christianmacion.github.io/portfolio/solutions/#eval-mcp-server tag:christianmacion26.github.io,2026:solution:eval-mcp-server Sun, 09 Aug 2026 03:08:15 GMT 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. MCP eval-harness multi-agent contract-first https://christianmacion.github.io/portfolio/solutions/#lookahead-bias-audit tag:christianmacion26.github.io,2026:solution:lookahead-bias-audit Sun, 09 Aug 2026 03:08:15 GMT 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. lookahead-bias point-in-time survivorship frozen-spec https://christianmacion.github.io/portfolio/solutions/#txn-cost-aware-backtest tag:christianmacion26.github.io,2026:solution:txn-cost-aware-backtest Sun, 09 Aug 2026 03:08:15 GMT 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. txn-costs slippage capacity walk-forward https://christianmacion.github.io/portfolio/projects/quant/09-lookahead-bias-audit/ tag:christianmacion26.github.io,2026:project:quant/09-lookahead-bias-audit Thu, 18 Jun 2026 00:00:00 GMT A leak inflates Sharpe 0.59 → 5.07; the shift test exposes it as 88% phantom. look-ahead-bias audit leakage methodology public-data https://christianmacion.github.io/portfolio/projects/quant/08-backtest-engine-costs/ tag:christianmacion26.github.io,2026:project:quant/08-backtest-engine-costs Thu, 28 May 2026 00:00:00 GMT High-turnover signal wins gross (0.20 > 0.13) but loses net of cost : break-even 20 bps. backtest-engine transaction-costs production-engineering public-data https://christianmacion.github.io/portfolio/projects/quant/07-macro-regime-overlay/ tag:christianmacion26.github.io,2026:project:quant/07-macro-regime-overlay Sun, 10 May 2026 00:00:00 GMT Vol-managed exposure: Sharpe 0.64 → 0.67, max DD −58% → −42%. regime overlay vol-targeting risk-management public-data https://christianmacion.github.io/portfolio/projects/quant/06-funding-carry/ tag:christianmacion26.github.io,2026:project:quant/06-funding-carry Sun, 05 Apr 2026 00:00:00 GMT Funding +11.9% annualized premium; fade IS 1.11 → OOS −0.05 (decayed post-2023). carry funding-rates crypto decay public-data https://christianmacion.github.io/portfolio/projects/ai/06-slop-scanner/ tag:christianmacion26.github.io,2026:project:ai/06-slop-scanner Sun, 22 Mar 2026 00:00:00 GMT 13-metric literature-grounded AI-output quality gate : drove a real draft from HEAVY (81) to CLEAN (3). evaluation content-quality nlg stdlib streamlit https://christianmacion.github.io/portfolio/projects/ai/05-reflect-revise/ tag:christianmacion26.github.io,2026:project:ai/05-reflect-revise Sun, 01 Mar 2026 00:00:00 GMT Reflection-loop agent : mean SLOP 127.5 → 14.0 across drafts, 3/4 improved, 1/4 honest no-progress halt. agents reflection evaluation halt logging https://christianmacion.github.io/portfolio/projects/quant/05-pairs-cointegration/ tag:christianmacion26.github.io,2026:project:quant/05-pairs-cointegration Wed, 18 Feb 2026 00:00:00 GMT ADF −2.11 (not cointegrated), half-life 208 d : the fade loses, as the test predicts. cointegration pairs-trading adf mean-reversion public-data https://christianmacion.github.io/portfolio/projects/ai/04-eval-mcp-server/ tag:christianmacion26.github.io,2026:project:ai/04-eval-mcp-server Wed, 04 Feb 2026 00:00:00 GMT MCP server exposing the slop-evaluation gate over Tools, Resources, and Prompts : 20/20 conformance, 100% round-trip parity. mcp agent-protocol evaluation tools resources prompts https://christianmacion.github.io/portfolio/projects/quant/04-variance-risk-premium/ tag:christianmacion26.github.io,2026:project:quant/04-variance-risk-premium Sun, 25 Jan 2026 00:00:00 GMT Implied > realized 85% of 36 yrs; predicts returns, Newey-West t = +6.5. volatility variance-risk-premium vix newey-west public-data https://christianmacion.github.io/portfolio/projects/ai/03-judge-harness/ tag:christianmacion26.github.io,2026:project:ai/03-judge-harness Mon, 12 Jan 2026 00:00:00 GMT LLM-as-judge pipeline validated against human raters : Cohen's κ = 0.58 with bootstrap CI and position-bias measured. llm-as-judge evaluation cohen-kappa bias validation https://christianmacion.github.io/portfolio/projects/quant/03-timeseries-momentum-voltarget/ tag:christianmacion26.github.io,2026:project:quant/03-timeseries-momentum-voltarget Sat, 20 Dec 2025 00:00:00 GMT Vol-targeting lifts Sharpe 0.27 → 0.39 and halves max DD (−62% → −30%). time-series-momentum vol-targeting risk-management public-data https://christianmacion.github.io/portfolio/projects/ai/02-toolcall-agent/ tag:christianmacion26.github.io,2026:project:ai/02-toolcall-agent Mon, 08 Dec 2025 00:00:00 GMT ReAct-style tool-calling agent with OTel traces, fault injection, and 100% tool/arg correctness. agents react function-calling opentelemetry fault-injection https://christianmacion.github.io/portfolio/projects/ai/01-rag-recall/ tag:christianmacion26.github.io,2026:project:ai/01-rag-recall Sat, 15 Nov 2025 00:00:00 GMT RAG service that proves its own retrieval : recall@3 = 0.886, MRR@3 = 0.805 with offline stdlib TF-IDF retriever. rag evaluation retrieval faithfulness stdlib https://christianmacion.github.io/portfolio/projects/quant/02-cross-sectional-momentum/ tag:christianmacion26.github.io,2026:project:quant/02-cross-sectional-momentum Wed, 15 Oct 2025 00:00:00 GMT IS Sharpe 0.91 → OOS −0.03 : an honest decay; bootstrap CI straddles zero. cross-sectional momentum oos decay public-data https://christianmacion.github.io/portfolio/projects/quant/01-deflated-sharpe/ tag:christianmacion26.github.io,2026:project:quant/01-deflated-sharpe Sat, 20 Sep 2025 00:00:00 GMT Best-of-160 BTC rule: IS Sharpe 1.14 is only 1.24× the pure-noise expectation : DSR = 0.70 (fail). multiple-testing deflated-sharpe rigor methodology public-data