Sentinel-X — AI-Powered Fraud Detection MLOps Pipeline (fraud classification training/evaluation)
Engineered a real-time fraud detection MLOps pipeline in which model training and evaluation required preparing labeled transaction datasets and running stratified cross-validation. Used MLflow experiment tracking to log performance metrics such as ROC-AUC, PR-AUC, MCC, and F1 for iterative model improvement and validation. Implemented an evaluation framework supporting production readiness decisions via model versioning and registry promotion. • Training/evaluation loops with metric logging • Stratified 5-fold cross-validation • Model versioning and registry promotion • Monitoring-driven retraining alerts
2023 - 2027