RA in the Stanford Vision and Learning Lab (Trustworthy AI group), advised by Fei Fei Li and Ehsan Adeli. I conducted research on ML interpretability and trustworthiness. We explored several research ideas, including 1) psychophysical experiments measuring the usefulness of interpretability techniques, 2) ideas for preventing neuron polysemanticity, and 3) developing a dataset decontamination benchmark for the task of decontaminating LLM datasets. Also, I co-authored the AI Safety and Alignment section of a landmark AI paper and was one of the voting authors to help coin the term "foundation model".

Research Fellow at Stanford Existential Risks Initiative, advised by Chelsea Voss (OpenAI) to do interpretability research on CLIP. I personally discovered 3 novel patterns related to how neurons contribute to other neurons using tools we built on top of the PyTorch Captum library.For my summer project (not advised by Chelsea Voss), I conducted economic growth theory research with the goal of developing an AI-aware and self-improvement-aware economic growth model.

TA for Logic, Problem-Solving, Group Theory, and Knot Theory at Stanford Pre-Collegiate Studies. Also, managed a team of five as head counselor. I was the youngest head counselor in the program's history (to my knowledge). Promoted from Residential Counselor (2018) to Head Counselor (2019).