About

All my work has been the same work: Making images. The instrument keeps changing.

With a camera, I was a photographer and cinematographer. I shot stills on the sets of independent films, made actor headshots, and designed movie posters for indie productions because I loved making them. With a whiteboard, I was a calculus teacher: AP Calculus BC, taught on camera every day when COVID closed the classrooms. Those lessons are still up at Richardson Calculus, untouched, because a classroom archive should stay a classroom archive.

Now the instrument is a model. I am a PhD candidate in data science at Kennesaw State University, with an M.S. in Analytics from Georgia Tech. I build diffusion models from scratch and study what they remember: When an image generator trains on private medical scans, what stays inside the weights, and who could find out? That question has pulled me into medical imaging, membership inference, and the strange interior of latent spaces.

The movie business never quite let go. These days I study Hollywood as a network: Hundreds of thousands of actors, every connection weighted by billing order, mapped for real influence rather than trivia. A designer who becomes a data scientist does not stop caring how things look. He starts asking how they connect.

Everything on this site follows one rule: If I cannot show it, I do not understand it yet. The essays run long, the figures are labored over, and the claims stay inside what was actually tested.

Find me on LinkedIn, or follow the papers on Google Scholar.