Projects
Some things I've built, written, or contributed to
-
Gone but Not Forgotten: Improved Benchmarks for Machine Unlearning
Alternative evaluation methods for machine unlearning algorithms — existing benchmarks rely on weak membership-inference attacks and overstate privacy gains. With Grimes, Abidi, Gallagher.
-
Out of Distribution Detection: Knowing When AI Doesn't Know
Three approaches to detecting when an AI model is operating outside its training distribution, framed as a reliability layer for safety-critical deployments.
-
Hands-On Computer Vision Tutorial
Notebooks and exercises for CMU-SEI's hands-on computer vision course, runnable via Binder or Colab.
-
Bias and Fairness in Modeling and Underwriting
A primer on algorithmic bias and model transparency for life-insurance underwriters.