I'm interested in GPU optimization — how we can optimize training and inference to exploit modern hardware features. I'm incredibly fortunate to have amazing mentors, such as Dan Fu and Chris Ré.
arXiv: 2410.01704. Explores a recent pluralistic alignment method to steer models towards preferences, comparing it against DPO and SFT across tasks. Research under Dr. Jan-Philipp Fränken @ Stanford AI Lab.
Won $10,000 in prizes at RunPod hackathon. Reduced GPU costs through self-healing spot clusters & FSDP
Received Outstanding Project Award from Prof. Tatsunori Hashimoto for CS 224N.
Acquired by Crimson Education in Sept. 2023. Searchable marketplace of 6K successful college applications that served 100K high school students. Undergraduate winner of Stanford BASES $100,000 prize and 2nd place winner at TigerLaunch 2023.
Grand Prize Winner & Meta Best Innovation Award at TreeHacks 2023 (largest college hackathon in the country w/ 1,800 participants). Mitigate overfishing through AI-augmented satellite imagery.