Hi! I’m Andy, an undergraduate Researcher and Engineer at UC Berkeley studying Electrical Engineering and Computer Science (EECS).

My research interests lie in building efficient and generalizable AI using Reinforcement Learning and interacting with the world through Robotics. Currently, I’m working on offline-to-online RL at Berkeley Artifical Intelligence Research, advised by professor Sergey Levine, and finetuning VLA foundation models with the GEAR team at NVIDIA. Outside my research, I enjoy building applied machine learning applications with my friends at Launchpad. I’m also currently interning with the ChipNeMo team at NVIDIA this summer, building LLM infrastructure.

If you’d like to chat, please reach out at [andypeng at berkeley dot edu].

Recent News

last updated: July 2025

• We released some exciting real world RL results with WSRL on a Franka robot! Check it out on the website
• I just released my first research paper with BAIR! We explore how to do RL fine-tuning without offline data retention. I’m excited to see how Warm-start RL (WSRL) can help scale up RL for real-world tasks, and the future of efficient RL finetuning. (Update: WSRL has been accepted to ICLR 2025!)