|
Vint Lee Hello! I'm a researcher at Amazon AGI SF, where I have the pleasure of working with Steven Zheng and Yifei Wang on fine-tuning and RL for coding agents. I also worked on browser-use agents previously. I completed my MS in EECS and BA in CS and Math at UC Berkeley, conducting Machine Learning research under the mentorship of Pieter Abbeel, Youngwoon Lee, and John Wawrzynek. |
|
ResearchI'm interested in deep learning in general, and post-training for agents in particular. In the past, I've also worked on model-based RL and diffusion models. |
|
Chip Placement with Diffusion Models
Vint Lee, Minh Nguyen, Leena Elzeiny, Chun Deng, Pieter Abbeel, John Wawrzynek ICML, 2025 code / arXiv We train diffusion models to perform macro placement, an important step in designing chips, using our novel data-generation algorithm and model architecture. |
|
DreamSmooth: Improving Model-based Reinforcement Learning via Reward Smoothing
Vint Lee, Pieter Abbeel, Youngwoon Lee ICLR, 2024 project page / code / arXiv Reward smoothing provides significant improvement to performance of model-based RL algorithms in many sparse-reward environments. |
|
Low-cost, Lightweight Electronic Flow Regulators for Throttling Liquid Rocket Engines
Vint Lee, Sohom Roy IAC, 2023. Best Undergraduate Paper Award project page / code / arXiv Electronic pressure regulators enable more flexible and accurate control of pressures in liquid rocket engines, opening up new capabilities for collegiate teams such as throttling. |
|
Website template from Jon Barron. |
