SHANKAR MANOHARAN

• Led the adoption of an in-house ML model for the debit portfolio, enhancing feature selection and optimizing model performance, and increased prevention by 5%.
• Boosted model performance by 0.5% to 1% through meticulous, periodic recalibrations.
• Executed comprehensive data extraction and preparation, directly contributing to a 3% increase in model performance.
• Engineered and optimized features, achieving a 1-3% performance uplift across multiple projects.
• Refined model evaluation criteria to enhance business alignment, boosting accuracy by 2%.
• Executed multiple iterations of ML models successfully deployed to production, driving over $5 million in benefits in terms of fraud and operational losses.
• Experimented with daily refreshes of ML models to dynamically adjust strategies using scalar parameters.
• Achieved an 80% reduction in score timeouts, decreasing from 0.25% to 0.05%, by enhancing real-time response efficiency.
• Provided direction and guidance to junior associates on complex tasks.

• Decommissioned one ML model and adopted another, evaluating models and redesigning entire authentication flow.
• Designed and implemented data flow pipelines for cleansing, preprocessing, and statistical analysis.
• Refined and automated reports using SAS and Python to enhance fraud prevention insights.
• Led two to three special card projects involving prepaid and single-use accounts, dedicating over 2,500 hours.
• Developed ROI-based fraud strategies, achieving annual savings exceeding $1 million.
• Streamlined client experience through strategic initiatives, resulting in a 3-basis points reduction in customer friction.

• Engineered comprehensive data pipeline architecture and implemented Python scripts for data flow utilizing AWS data pipelines with Boto3 for diverse data science projects.

• Re-hosted & optimized tax modules to fit the target system (Java) by understanding the target architecture

• Streamlined claim processes and recommended solutions for various business issues which saved 20% support hours

• Involved in functionally structuring skeletal modules which avoided 150 defects being created by testing team

• Worked on SQL queries, cursors, and performance tuning by reducing retrieval time by 20 secs per query

• Engineered a new database on DB2 which involved indexes, triggers and table partitioning for QA team of 80

• Designed interface programs in Cobol & Java for data flow of supplies from Mainframe to 3rd party vendor systems