Background

I started in computational physics — not because I wanted to study physics, but because physics forces you to think in systems. Mass is conserved. Momentum has direction. The constraints are real and non-negotiable. That discipline stays with me: before building anything, I want to understand what's actually invariant and what's just convention masquerading as constraint.

From there I spent several years in a defense contractor proposal shop, which sounds unrelated but wasn't. My job was to translate between what engineers could build and what decision-makers could act on. The gap between those two things is enormous, and closing it is mostly a problem of systems design — not communication style. A system that requires people to change how they work in order to use it isn't a solution; it's a bet that your users will do the harder job so you don't have to.

AI engineering is where both threads converge. The tools are powerful enough that the limiting factor is almost never capability — it's whether the system fits into how people actually operate. I build at that boundary. The goal isn't impressive demos; it's adoption without a training session.