Alekhya A

• Developed XGBoost trade failure prediction model using Python, SQL, Snowflake, and SHAP, reducing settlement failures 27% and saving $4.2M annually. • Automated reconciliation workflows with Airflow and AWS S3 pipelines, improving Straight Through Processing from 91% to 96% and reducing manual workload 38%. • Engineered an LSTM and XGBoost ensemble liquidity forecasting solution in PyTorch and Databricks, reducing MAPE 34% and improving liquidity forecasting accuracy, enabling better capital allocation decisions. • Architected real-time treasury analytics platform using Spark, AWS Lambda, and REST APIs, automating 75% forecasting workflows with sub-two-second dashboard latency. • Designed graph-based AML detection system leveraging PyTorch Geometric, Neo4j, and Kafka, reducing false positives 41% and increasing suspicious detection 29%. • Implemented enterprise-scale ML deployment using Kubernetes and MLflow, ensuring model governance, audit traceability, and regulatory compliance under OCC standards. • Optimized capital and liquidity risk management strategies aligning Basel III LCR requirements, improving compliance buffer efficiency 22% across multi-currency institutional portfolios

• Processed and cleaned large financial datasets using Python, SQL, and PySpark, improving data quality and consistency to support downstream predictive modeling and reporting workflows. • Built and evaluated machine learning models using Scikit-learn, XGBoost, and LightGBM to predict loan default risk, improving classification accuracy by 10–12% and reducing false positives. • Contributed to developing ETL workflows using Apache Airflow and Spark, reducing manual processing efforts by 2025% and improving pipeline execution efficiency by 15–20%. • Applied feature engineering techniques including behavioral segmentation, payment trend analysis, and statistical aggregations to enhance risk scoring models and improve overall prediction stability. • Assisted in deploying ML models using Docker, Flask, and AWS services (EC2, S3, Lambda), helping reduce inference latency by 15–20% and improve deployment consistency. • Monitored model performance and data drift using MLflow dashboards, supporting retraining cycles and improving production model stability across financial risk prediction systems. • Collaborated with data engineering and DevOps teams to support ML feature releases, improving reporting turnaround time by 15–20% and enhancing cross-team delivery coordination.