Lichang Chen

I am a research scientist at Google DeepMind, building state-of-the-art coding agents to solve challenging problems in physics. Previously, I worked at Meta Superintelligence Labs on agents for automating AI research and self-improvement through reinforcement learning and continual learning. I am the main contributor to the SWE coding capabilities of Muse Spark 1.0, 1.1, 1.2, 1.3, and Muse Wonder. I did my PhD at the Computer Science Department, University of Maryland, College Park. I can be reached at {bob}{my-first-name+my-last-name}@gmail.com.

My name in Chinese: 陈力畅

Google Scholar  /  Twitter  /  LinkedIn  /  Github

My research interests are in reinforcement learning, self-improving coding agents, and AI for scientific discovery. My current focus is building coding agents that can tackle challenging physics problems and accelerate scientific research.

Research Scientist @ Google DeepMind, Present. Building coding agents to solve challenging problems in physics.
AI Research Scientist @ Meta Superintelligence Labs, Previously. Agents for automating AI research.

Intern@Google Deepmind, Meta RL for reasoning & Thinking to Learn.
Intern@Google Deepmind, The evaluation/alignment of Omni-Modality Language Models.
Intern@Google Research & Cloud AI Research The self-improvement of multimodal LLMs.
Intern@NVIDIA ADLR team, Mitigate hackings in RLHF/More Robust Reward Models