Featured
An educational stock analysis tool that keeps the decision out of the language model. A pure, rule-based evaluator produces the signal from technical indicators, and the LLM only writes the explanation, so the same inputs always give the same result.
Layered design (data, decision, narrative, UI), multi-provider LLM fallback, 16 tests, and CI across Python 3.10 to 3.12.
Python, pandas, Streamlit, LangChain (Anthropic, OpenAI, Mistral), pytest, GitHub Actions
For learning only. Not financial advice.
Agent audit and governance
A tamper-evident audit trail for an AI agent gateway: per-tenant hash chains with scheduled integrity verification and alerting. Running continuously on self-hosted hardware since August 2026.
TypeScript, Python, SQLite, Vitest
Private repository. Happy to walk through it in an interview.
Second Brain
A goal and task execution system built around the Buffett 5/25 rule, with a CLI, a REST API, and a React UI. Claude ranks open work and links tasks to goals across planning horizons.
Python, FastAPI, React, TypeScript, Anthropic API
Private repository. Happy to walk through it in an interview.