Work

Estimated the causal effect of Washington State’s 7% capital gains tax on the out-migration of high-income households in 2021. Applied a differences-in-differences design, constructing a synthetic control unit matched on Washington’s migration pre-trends for low-income households. Using a placebo permutation test, found statistical evidence that the tax caused a moderate out-migration of high-income households.

Created an informal mathematical statement dependency graph consisting of 11.7 million statments connected with over 18.3 million dependencies as part of the Universiy of Washington’s Math AI Lab. Our goal was to create the first informal-formal mathematical dependency graph, connecting informal mathematics (i.e. arXiv papers) with formal mathematics (i.e. Lean libraries). Led development of informal statement parsing and dependency connecting, creating a pipeline using AWS services, Nebius Token Factory, and UW’s Hyak HPC cluster for under $5,000.

Built a semantic search engine for mathematical theorems as part of the University of Washington’s Math AI Lab. Led development of the world’s largest theorem database (9.2 million) using AWS services and a Python pipeline. Accepted to the 2026 ICLR Workshop in LLM Reasoning.

Analyzed how well the Mankiw-Romer-Weil Solow growth model captures the economic dynamics of the U.S., Japan, and Germany from 1950–2023 with Python. Found that a SINDy-driven model outperformed those with constant or time-varying exogenous parameters.

Examined how Okun’s Law changed in the U.S. across recession periods and states using 2005–2025 data with R. Found no evidence of heterogeneity across periods, and only weak evidence across states.