Caleb Goodman

I care about software that actually works — not just ships. There's something genuinely satisfying about building systems where correctness is load-bearing, where a wrong assumption doesn't just produce a bad user experience but breaks something real. That's what drew me to software engineering in the first place: the discipline of it. The craft of turning a fuzzy problem into something precise, testable, and maintainable.

More recently, that same instinct has pulled me toward computer vision — specifically the parts of it that most people don't talk about. Not object detection, not segmentation. I'm interested in the forensic side: sensor fingerprinting via Photo-Response Non-Uniformity (PRNU), photometric integrity analysis, anti-spoofing and passive liveness detection, and the detection of video injection attacks in biometric pipelines. These are areas where the adversary is sophisticated and the margin for error is essentially zero. I'm involved in research at UIUC that keeps me close to the literature. If you want to go deeper, I've linked suggested reading at the bottom of this page.

I cold-outreached Dyneti earlier this year and ended up joining the team. I build models there. I'd rather not get into the specifics — that work will speak for itself in time.