Skip to main content

Data Science

I build choice models and simulations that predict how customers respond to incentives, with a focus on energy and transportation. My main project, on electric vehicle smart charging, is a partnership with Southern California Edison, Pacific Gas & Electric, and National Grid. I ran a choice experiment with 1,356 owners to model what makes them enroll, then simulated 10,000 vehicles on the grid to find the most cost-efficient level of enrollment. My methods also include optimization, agent-based and discrete-event simulation, system dynamics, and network analysis.

Open-Source Software

I develop open-source R tools for survey design and discrete choice modeling. The main one is surveydown, a programmable, markdown-based survey platform built on Quarto, Shiny, and PostgreSQL, with sdstudio as its companion GUI. surveydown has more than 180 GitHub stars and 11,000 CRAN downloads, and its methods paper is published in PLOS ONE.

Agentic AI Engineering

I build software with agentic workflows around Claude Code and Codex, which I extend with skills, MCP servers, and hooks. On the blog, I write tutorials and share my own workflows, and I ran a workshop at GW TAI for researchers and developers.

With Professor Helveston, I co-founded Alkaline Software, an app development studio where I am the principal engineer. We build desktop apps in Rust with GPUI, using the same workflows. I lead development of our main app, Emaki, an Agentic Development Environment (ADE), and I built the studio’s app template for macOS, Windows, and Linux, as well as its design system.