Hi, I’m Michael!

I’m a researcher at OpenAI, working on RSI metrics and broadly interested in understanding the future of AI development and its impact on research and society. I’m based in San Francisco, CA.

Previously, I worked on code mid-training, post-training, and evals at Google DeepMind, working on Google Antigravity and Gemini 3 to 3.8. Before that, I was on the model training team at Windsurf (reverse acquihired to GDM).

I graduated summa cum laude and Phi Beta Kappa from Harvard College with an A.B. in Computer Science and Government, and also concurrently completed my A.M. in Statistics. On campus, I led Harvard Computer Society, organized study groups with visiting fellows from Congress and the Cabinet through the Institute of Politics, sang all four years with the Harvard Glee Club, and published computational social science research on redistricting with Prof. Kosuke Imai’s ALARM Lab. My senior thesis “Words Speak as Loudly as Actions: Deep Learning Methods for Stance-Based Ideal Points from Congressional Speeches,” advised by Prof. Ariel Procaccia and António Câmara, won the Thomas Temple Hoopes Prize for outstanding thesis and the Gerda Richards Crosby Prize for the highest-ranked student in Government.

I love pursuing my hobbies in the Bay Area—you can currently find me working on my private pilot’s license at KSQL and singing with the San Francisco Choral Society and Masterworks Chorale. I’m always happy to meet new people, gain new experiences, and chat about big ideas and interesting tidbits in the world. Feel free to send me an email at michaelzhao[at]alumni[dot]harvard[dot]edu, or find me on X!