research

ml + systems @ stanford

hazy research, sail

working on an embedded CUDA DSL to accelerate inference and training for transformer-based models on h100s

JRT-arXiv, LolCats-arXiv, TK-arXiv

ml + systems @ stanford

stanford graphics lab

co-optimizing the design and execution of ml architectures that enables end-to-end (raw pixel inputs, agent action outputs) rl algorithms to quickly learn effective game-playing policies on the madrona game engine

code

rl @ stanford

autonomous systems lab, sail

building scalable & robust rl systems that utilize graph convolutional networks to improve the control and operation of electric autonomous mobility on demand systems

arXiv, code

robotics @ umich, ann arbor

precision systems design lab

designed a novel 2-dof parallel kinematic prosthetic wrist and a high-performance, low-weight prosthetic gripper

paper