Fancy perturbation prediction methods (e.g. STATE, GEARS, diffusion models) don’t readily outperform FM embeddings + simple predictors (e.g. kNN). (March 29, 2026)
scRNA-seq FMs have room to improve, but they have moved beyond linear baselines. (March 27, 2026)
Manual CPU and GPU management on a shared machine. (August 16, 2025)
Breaking down how all of the parameters in a vision transformer are used. (September 1, 2024)
If I could go back in time to the start of my Ph.D., these are a few of the books I would make myself read. (January 1, 2024)
This summer I will be moving to San Diego to join Altos Labs as a Machine Learning Research Scientist. (May 5, 2023)