Igor Geyn

@roguski.bsky.social

168 Followers 58 Following

Independent researcher and data scientist (ex: PG&E, BGOV, others) working on causal inference, machine learning, and estimating policy effects in different domains. PhD from @UCLA

Chat interface titled "Ask CalBallot." A user asks "What were the 5 closest ballot measures in the last 10 years?" and the bot responds with a SQL query and a results table showing five measures decided by margins under 1%, topped by a 2016 Fresno sales tax that passed by 0.01%.

Campaign finance view for Prop 36 showing side-by-side Support ($12.4M, led by Walmart and CA Grocers Assn) and Oppose ($3.1M, led by CA Teachers Assn and ACLU) panels with top donors, a fundraising timeline bar chart, and a donor composition bar with a grassroots score of 8%.

Detail page for Prop 36: Criminal Penalties and Treatment (2024, Passed 68.4%). Shows an AI-generated plain-language summary, a vote bar, and a grid of four related measures found by text similarity.

Heatmap table with 11 county rows (Statewide, Los Angeles, San Francisco, Alameda, San Diego, Santa Clara, Contra Costa, Orange, Sacramento, San Mateo, Fresno) and 4 topic columns (Tax, Education, Housing, Safety). Cells are color-coded from green (high pass rate) to red (low), with pass percentages and measure counts in each cell.

Year filter panel showing filter buttons for level, region, topic, year, and status, with the year filter expanded to display clickable chips for every election year from 2026 down to 1998, each showing a measure count.

CalBallot homepage showing a dark header, stats ribbon (11,485 total measures, 67.5% pass rate, 1998–2026), filter buttons for level, region, topic, year, and status, and a 2x2 grid of measure cards including Prop 36, Prop 3, Prop 33, and Prop 32.