At Down, I ran the data side of about 90 A/B tests and built two trust-and-safety models: one to detect bad actors from their first five minutes of behavior, and another to identify underage users. Together, the models cut manual review from 6.5% to 4.5% of new accounts.
I also led Down’s move to dbt and set up daily monitoring to catch tracking losses and data gaps before they affected experiment readouts. Across fintech, gaming, and dating, I’ve built machine-learning models, automated reporting that saved more than 20 hours a week, and turned analysis into clear recommendations.
