About

I build systems where the decision has to be explainable — agents that route across real tools, and scoring pipelines that affect real people.

Two things I’ve published:

• langgraph-code-agent — a local, permissioned autonomous agent on LangGraph. Cascade tool router across 14 MCP servers, validated on a sealed hold-out set.
• resume-screening-pipeline — local-first resume screening where every score is deterministic and decomposable, and no email sends itself.

Find me on GitHub.