Data Scientist
Feb 2023 – Present- Built a gradient-boosted churn model in XGBoost scoring 1.4M subscribers weekly, enabling save offers that retained $2.1M in annual recurring revenue
- Designed 14 A/B tests with product managers, using CUPED variance reduction to cut required sample sizes by 30%
- Productionized models with MLflow and Airflow, reducing retraining time from 2 days to 3 hours
- Partnered with finance to forecast quarterly subscriber counts within 2.5% error using Prophet and hierarchical reconciliation