• Engineered ML-based credit scoring models using Logistic Regression, Random Forest, Gradient Boosting, and CIBIL/Experian/Equifax APIs, processing 1M+ loan applications per month, reducing loan default rates by 27% and improving approval speed by 60%.
• Developed advanced behavioral feature engineering pipelines using Python, Pandas, PySpark, UPI transaction analytics, and Account Aggregator (AA) Framework, increasing approval rates for new-to-credit customers by 22%.
• Implemented explainable AI models using SHAP, StatsModels, and Python, ensuring 100% RBI compliance for model transparency and credit risk reporting.
• Designed low-latency fraud detection pipelines using Isolation Forest, Autoencoders, Apache Kafka, Spark Streaming, and NPCI/Payment Gateway APIs, detecting 92% of fraudulent transactions in milliseconds while reducing false positives by 35%.
• Orchestrated cloud-native ML workflows using AWS S3, EC2, Lambda, Docker, Kubernetes, MLflow, and Jenkins, reducing end-to-end model deployment time by 50% and supporting 20M+ transactions/day.