Most of what I built last year is AI infrastructure.

I'm a software engineer in Atlanta. My title says software engineer; most of what I've built in the last year is AI infrastructure. That is the transition this site documents, from writing the software to building the systems that write it.

The constraints do not negotiate.

Healthcare is where I learned to build under constraints that don't negotiate: data that can't leave the VPC, releases that reach 61 production instances, a platform carried through an acquisition. That's a tool in the box, not a category I want to live in.

The paper I point to has aged well.

Before this: a Georgia Tech MSCS, a stint teaching health informatics as a TA, and four published papers. The one I'd point you at isMachine Unlearning: Its Need and Implementation Strategies, on how you remove a person's influence from a model after they withdraw consent. It has aged well.