Sharanya V

• Advanced credit underwriting analytics for $2.5B+ consumer lending portfolio by constructing gradient boosting and deep neural network models in Python and PyTorch, elevating risk segmentation lift by 21% across 10M+ applicant records. • Orchestrated large-scale data preparation workflows within Snowflake and Databricks, refining 18M+ cardholder and transaction records, reducing feature computation time by 33% and strengthening decision latency for real-time authorization systems. • Productionized model scoring pipelines using FastAPI, Kubernetes, and AWS SageMaker endpoints, sustaining 75K+ daily credit decision requests under 250ms latency while improving digital approval conversion rates by 16%. • Designed improved fraud detection frameworks leveraging graph analytics and anomaly detection techniques on 22M+ payment events, decreasing fraudulent exposure by 14% and lowering false-positive customer declines across retail banking channels. • Instituted continuous model governance controls integrating MLflow tracking and eliminated manual intervention in drift diagnostics across 8 live risk models, curbing performance variance by 26% and reinforcing SR 11-7 regulatory compliance standards. • Partnered with product strategy and risk operations leaders to convert 30+ policy directives into predictive scoring enhancements, enabling $48M incremental revenue protection through accelerated credit line assignment and portfolio management decisions. • Re-architected distributed batch and streaming inference workflows using Apache Kafka and Spark Structured Streaming, processing 6M+ daily account activities, boosting processing throughput by 38% while preserving enterprise SLA adherence.

• Engineered supervised learning models using Python, scikit-learn, and XGBoost on 8M+ banking transactions, improving credit risk classification accuracy by 17% and reducing default exposure across two retail lending portfolios. • Built end-to-end ETL pipelines leveraging SQL, Pandas, and Apache Spark to cleanse and transform 12M+ customer records, decreasing data latency by 28% and enabling near real-time underwriting decision support. • Deployed REST-based ML inference services using Flask and Docker, supporting 50K+ monthly API calls with sub-300ms response time, increasing loan processing throughput by 22% across digital banking channels. • Conducted feature engineering and model validation across 40+ financial variables, applying cross-validation and ROC-AUC evaluation techniques, improving fraud detection precision by 19% and minimizing false-positive transaction blocks. • Automated model performance monitoring using Python scripts and SQL dashboards, tracking drift metrics across 6 production models, reducing performance degradation incidents by 24% through initiative-taking retraining cycles. • Collaborated with business analysts and compliance teams to translate 25+ regulatory reporting requirements into ML-driven analytics workflows, enhancing audit traceability and strengthening regulatory adherence across quarterly submissions. • Optimized batch scoring processes on Hadoop clusters processing 5M+ daily records, improving compute efficiency by 31% and lowering infrastructure costs while maintaining SLA compliance for enterprise banking clients.