At Amazon, I build real-time XGBoost models that score accounts across money-laundering typologies. I handle model development from investigation-outcome labels and feature engineering through validation and model-risk documentation.
My work raised precision by 75% and cut false positives by 80% compared with the previous rule-based detection. I also added SHAP reason codes for investigators and auditors, and started a GenAI project using agentic workflows that cut analytical pipeline build time by 85%.
