About

I build production AI systems and the proof around them. Not demos — platforms, playbooks, and verification layers that hold up when someone is depending on them.

Before any of that I spent a decade on a warehouse floor, running inventory, materials, and fleet. That is where I learned that a system is only real if it works on the worst shift of the week, with the tired crew, under a deadline. Every product I build now is designed against that standard.

I treat LLMs as brilliant but scattered collaborators. Reliability does not come from a clever phrase — it comes from architecture: prompt systems, multi-agent orchestration, evaluation, and human-in-the-loop gates. I teach that craft because most teams have been burned by generic prompting and need habits that survive real work.