Building data-driven systems that turn complex signals into real-world predictions.
Developed Voter DNA: a machine learning model trained on 60,000+ samples that simulates how demographic traits and their interactions shape voting behavior. Combines LASSO regression with an interactive interface for real-time voter profile prediction.
Also building OSZ Polls, a platform for aggregating U.S. polling data and modeling district-level election outcomes through live, map-based visualizations.
• Built 90% of the platform end-to-end (front-end → data pipelines)
• Developed a Swingometer to simulate U.S. House elections
• Designed district-level dashboards with live projections
• Integrated real-time polling APIs for dynamic visualization