Vikram Kumar

Collection and Fraud Analytics:

• Developed and deployed a recovery model using XGBoost to predict the likelihood of recovering amounts from previously written-off customers, integrating on-us and bureau data, performing advanced feature engineering, and optimizing the collection process for targeted recovery actions, enabling cost-effective collections, and strategic alignment with business objectives.
• Created a bureau-based risk categorization framework in the Fraud Analytics team to classify customers into high, medium, and low-risk segments, improving early fraud detection, and improving overall risk management efficiency.
• Designed and implemented a lead score cutoff framework leveraging vendor scorecards to proactively reject low-quality, high-fraud risk leads, improving fraud prevention and reducing approval turnaround time by 20%.
• Developed and implemented automated Early Warning Signal (EWS) triggers at both the branch and customer levels, improving early risk detection, and enabling the FCU team to proactively monitor sourcing quality and reducing potential defaults.

Portfolio Analytics:

• Conducted an analysis on the CRIF INDV score equivalence with the CIBIL score for a product migration, enabling the development of BRE (Business Rule Engine) rules to ensure a seamless bureau transition and maintain consistent risk assessment.
• Performed reject analysis using bureau data to identify lost opportunities from rejected leads that were later booked in the market, uncovering performance trends, and targeting segments for improved customer onboarding in the future.
• Conducted an in-depth analysis of demographics, on-us, off-us performance to identify optimal locations for branch expansion, focusing on rural areas, and contributed to the successful opening of branches at 85 new locations, driving significant business growth.