About

I build AI agent systems and help teams get them into production.

Right now that means a retrieval pipeline built on LangChain and pgvector that contact center agents and supervisors at a credit union use every day; an MCP server, in development, that will give that client's internal AI teams one governed way into the same call data; and a multi-agent content system with LLM-as-judge evaluation, guardrails, and a human review stage. I build with coding agents, mostly Claude Code and Pi, and a good part of the work is catching what they get wrong.

Before this I spent most of my career in field SE and SA roles at IBM, Microsoft, and AWS, and at a few startups, working out whether a product actually solved a customer's problem and helping them decide what to do about it. That's still the part of the work I like best, and it's why the posts here cover the tradeoffs and the rough edges along with what got built.

The Anchoring AI series walks through a portfolio version of one of these projects in detail.

I'm in San Antonio.