shenn667
AI Systems & Tooling Engineer — deterministic tooling for AI systems: context infrastructure, eval gates, verifiable RAG.
- cited-rag — Evaluation (Software Quality Engineering)
Reproducible benchmark (measured on v0.1.0, identical on v0.1.1): Recall@5 = 1.00, MRR = 0.93, correct-refusal = 1.00 (12/12), n = 52. Protocol, gates C1–C6, honest limitations.
- cited-rag on GitHub
Offline hybrid RAG (BM25 + TF-IDF + RRF) with citations. Stdlib-only.
- cited-rag on PyPI
v0.1.1 (70 tests) — pip install cited-rag.
- agent-context-bridge (MCP Server)
v0.3.2 (92 tests, -98.7% tokens). GitHub/PyPI/MCP Registry/Glama. Context persistence for AI agents.
- llm-eval-ci (CI Gate)
v0.1.2 (159 tests, 100% check coverage + check-text CLI). Deterministic evals that fail builds on regressions.
- claude-verified-agent (Verified Agent Demo)
48 tests, 4 CI gates + Banking/SecOps vertical. End-to-end: context → cited answer → PII/injection gate.
Static site for shenn667.github.io. See /eval/ for the full SQE report.