I am currently building a stealth startup. My interests lie broadly in building next generation AI systems and models (e.g. efficient serving and post-training). Previously, I was at Google DeepMind, where I worked on efficient agentic systems for serving and post-training.

I created the Agentica Project, an open-source initiative for post-training language agents via RL, producing models like DeepScaleR, DeepCoder, and DeepSWE with over 2M+ downloads. Our training recipes are published on the rLLM framework.

I received my PhD in EECS from UC Berkeley, where I was advised by Ion Stoica and was part of SkyLab and BAIR. I also hold an M.S. and B.S. (CS & Business) from UC Berkeley.

For a full list, see my publications page or Google Scholar.

An open-source initiative for post-training language agents via reinforcement learning. Agentica models have achieved over 2M+ downloads on HuggingFace. We've optimized RL systems and published training recipes on the rLLM project, now maintained by academia, with over 5K+ stars.

University of California, Berkeley
Ph.D. in Electrical Engineering and Computer Science, Aug 2021 - Fall 2025
Advised by Ion Stoica

University of California, Berkeley
M.S. in Electrical Engineering and Computer Science, Aug 2020 - May 2021
Advised by Ion Stoica and Ken Goldberg

University of California, Berkeley
B.S. in EECS and Business Administration (Summa Cum Laude), Aug 2016 - May 2020
GPA: 3.98/4