I like robot learning, math, automated reasoning, CS theory, and a few other things.
Research at The Movement Lab. Currently interested in whole-body locomotion for humanoids and RL post-training for dexterous manipulation.
Projects
- llm-projectsLanguage model coursework: tokenization, training systems, scaling, data processing, and reasoning.
- taso-sparseAutomatically deriving neural network compiler optimizations from mathematical identities.
- PiTorrentPeer-to-peer file sharing over Bluetooth on bare-metal Raspberry Pis.
- ChokoPPOReinforcement learning agents for Choko using PPO, Double DQN, and self-play.