Hello! I am an undergraduate and master's student at Stanford studying computer science and math.

I build predictive models of complex systems from messy, open-ended data. In the social sciences, I've focused on simulating human behavior. At Stanford with Michael Bernstein, I curated large-scale behavioral datasets and post-trained language models to be representative and calibrated for these tasks. As a researcher at Simile, I scaled this method and deployed simulations for market researchers.

I'm increasingly convinced this machinery matters most where data is scarce and mistakes are expensive; you can't A/B test an airplane or ship an "MVP" of a bridge. Learned surrogates change what's tractable in these domains, and the hard problems (e.g., validation, model fidelity, scalability, uncertainty quantification) carry over.

If you are interested in chatting about any of this, please reach out!