I teach machines to see.

For the last 10 years I've worked on one question in four different industries: how do you get a camera to understand what it's looking at — and act on it, reliably, at the edge, in milliseconds?

The answers have shipped. At Aptiv, the interior sensing systems I built ride along in BMWs, GMs, and Fords — reading gestures and making sure drivers watch the road. At Ford, I researched the driver monitoring behind BlueCruise. At Rotor, I built the perception pipeline for an autonomous helicopter. And at Sam's Club, I helped scale the AI-powered exit from pilot to 600 clubs across the US — letting members skip the receipt line entirely.

The gap between a paper and a product is where I like to work.

What keeps me hooked is the unglamorous last mile: quantizing a model until it fits a vehicle-grade chip, profiling a pipeline until it hits frame rate, and designing for the one-in-a-million frame where getting it wrong actually matters. Off the clock, that curiosity spills into ray tracers, camera ISPs, and neural rendering experiments.