Lately I've been building — open-source guardrails that make AI agents earn trust before they get autonomy. It's live on PyPI. Before that, I built so my own agents could trade through a real brokerage.
Upholdr grants agents graduated autonomy per risk category — report_only, suggest, then auto — and backs every action with a tamper-evident, hash-chained audit ledger you can replay deterministically. Fail-closed by default. Live on PyPI · GitHub
fintools-mcp is an MCP server that gives AI assistants live market data and brokerage tooling — options chains, quotes, and order workflows — built to run a real systematic trading operation, not a demo. GitHub
Simtek is where I've gone deepest on agents — a live, broker-connected trading system with a governor agent that's earning its autonomy in stages. Its shadow phase caught a failure that would have burned it on day one; I wrote that story up in My agent would have been wrong on day one.
I studied Law & Policy at the University of Washington, picked up a Python certificate there, and taught myself the rest — everything above is self-built, shipped, and public. I'm in Tacoma, WA, open to remote engineering roles. The fastest way to evaluate me is to read the code.
