Embedded, technical,
and measured.

I build AI workflows and internal tooling for the people doing repetitive work, then stay close enough to know whether they actually use it. My default is to sit with the team first, prototype in days rather than weeks, and put something real in front of them before anyone writes a spec.

Seven years in software quality shaped how I build. My instinct with any system is to ask where it breaks and how anyone would know. That matters more with AI, not less, because the output is probabilistic and the regressions are quiet. I build the evaluation layer alongside the thing itself.

What I care about is whether the workflow is still running on Monday morning, and whether the team can change it without me.