I'm a senior at UC San Diego studying computer science with a focus on machine learning and systems. I'm interested in pretraining research for foundation language and world models, scaling RL for physical AI, AI safety and interpretability, and efficient inference.
I'm currently doing research with the Hao AI Lab and NVIDIA Research on pretraining and efficient inference for foundation world models, and building FastVideo -- , an open-source framework for accelerating video generation through efficient post-training, distillation, and serving of large-scale generative models.
I'm also exploring how model compression techniques interact with learned safety behaviors in large language models. Learn more here.
This past fall, I took some time off from classes to intern with the Autopilot team at Tesla AI, where I got to work on modeling and GPU kernels for SOTA 3D foundation models powering FSD and Optimus. Prior to that, I was on the PyTorch team at Meta, where I helped build out the software acceleration stack for next-generation training and inference chips.