At Mutt Data, working with Coca-Cola FEMSA, I built a production causal evaluation engine that estimates incremental sales from commercial initiatives every two weeks for 1.5M+ customers in Mexico and Brazil. It compares Matrix Completion for Causal Panels with Causal Forest and RT-Learner benchmarks.
I designed and analyzed 10 controlled experiments across Mexico, Brazil and Colombia, defining treatment and control groups, sample sizes, statistical power and randomization. I presented results to business teams, including findings on a recommendation program and app adoption.
I develop transaction-data pipelines and models with Python, PySpark, Spark SQL and Databricks. I also estimate sales effects using difference-in-differences, synthetic controls and LightGBM-based causal models.
Previously, at Banco Santander, I developed XGBoost classification and time-series forecasting models for banking risk. At Kevin D. Oden & Associates, I validated predictive and decision models for banks through backtesting, sensitivity analysis and challenger models, and designed monitoring frameworks.
