Gary Wang
Hello everyone! I study statistics, data science and economics at UC Berkeley. I work on machine-learning evaluation research and build software for data, markets, and simulation.
University of California, Berkeley
Bachelor's degrees · Statistics, Data Science & Economics
Statistics, data science and economics at UC Berkeley. My research keeps returning to one question — how do you tell whether a machine-learning system is actually good — and most of what I build is an attempt to answer it in public.
Right now that means AI nutrition and disease-risk research through Berkeley's URAP, AI for science discovery with the Data Science Discovery Program, AI peer-review evaluation at Rapid Reviews with the MIT Press and the Gates Foundation, and social-media analysis at the Leonard Transportation Center. Before that: tensor methods for chatbot evaluation at UC Riverside, reward modeling and faithfulness evaluation at the Leonard Transportation Center, and econometric measurement of AI-driven commerce at Alibaba. Alongside coursework I design and ship software — evaluation tooling, market scanners, simulations — because a system you have deployed is harder to fool yourself about than one you have only described.
AI nutrition and disease-risk research and AI for science discovery at Berkeley, and AI peer-review evaluation at Rapid Reviews with the MIT Press and the Gates Foundation. Before that, tensor and CP decomposition for classifying chatbot responses at UC Riverside, and reward modeling at CSU San Bernardino and the Leonard Transportation Center. Earlier: an NLP paper at ADNLP 2024 and a U.S. patent, issued May 2024.
Kinetic
A camera-only hand-tracking game controller — your hands become the gamepad, no hardware.