I build image models from scratch, and I study what they remember.

PhD candidate in data science at Kennesaw State University, M.S. in Analytics from Georgia Tech. Former photographer and cinematographer, former calculus teacher. I study what image models remember, I map Hollywood as a network, and I explain both with pictures.

Work
How Diffusion Models Memorize
A diffusion model trained from scratch on 300 photographs from one film set, and the membership inference attack that finds them again. Scroll to watch it learn, then run the attack yourself.
Interactive · 2026
All work →
Essays
Preseason: Simulating a Million College Pick'em Entries Across 10,000 Seasons
An unreasonable amount of machine learning, aimed at a free football Pick'em contest. Ten thousand simulated seasons, a million simulated opponents, and a year-long attempt to out-forecast Las Vegas. Part one: What the game is, and what the numbers say it takes to win it.
Sports Analytics · 2026
All essays →
Selected publications
SMITE: Step-wise Memorization Inference with Two-channel Error for Membership Inference on Fine-Tuned Retinal Diffusion Models
Richardson, Aygun, Karakaya · IMNS 2026 · to appear
Six Degrees of Kevin Bacon and the Weighted Hollywood Actor Network
Richardson, DeMaio · CAC 2026 · to appear
Benchmarking the Dell Pro Max GB10 Against the NVIDIA DGX-1
Richardson, Aygun, Karakaya · IMNS 2026 · to appear
Image Quality Assessment of VistaView & iNview Portable Retinal Imaging Systems
Richardson, Suer, Aygun, Karakaya · IEEE EMBC 2026 · to appear
All publications →