Utkarsh Vaibhav

Took a career break to reassess personal priorities and spend more time with family

• Designed & helped implement a complete personal loan cross sell analytics platform from scratch for an automotive lender
• Oversaw the development of a CNN based credit risk model to predict customer's intent to pay using Bureau, SMS and alternative data. Worked on the strategy to reduce risk and increase approval rate
• Led as a functional expert in suggesting & deriving the growth KPIs to the existing dashboards across departments
• Developed a ML system using RAG to detect threatening or abusive language used by agents during collection calls, enabling proactive measures to reduce customer escalations

• Amazon Reviews Analysis using NLP: Used NLP for topic modelling and separated good reviews from bad ones using clustering from a database containing close to 30000 reviews of competitors' listings
• This analysis helped us discover what product features do the customers most resonate with and offer them that
• Achieved a recall of 75%

• Category Recommendation Engine: Ideated and led the project to predict customers spend on various categories on competitors card for low engaged/inactive customers and used it to formulate an on-us engagement and loyalty strategy. Achieved a 95% accuracy for competitors’ spend for top 30% spenders
• Program Evaluation and Risk Assessment: Predicted the NPV of a prospect by building a Credit card PnL for both proprietary and co-branded Citi credit cards using SAS and Excel to target the most profitable prospects with an 85% accuracy

• Application & Behaviour Scorecard for Credit Card Portfolio: Designed the acquisition approval cut-off strategy based on best model parameters using a Logistic Regression model
• Optimized approval rates w.r.t risk appetite that helped approve 73% of applications, reducing 9.4% bad customers to 4.2%
• Cross Holding Data model: Developed a data model at customer level showing various banking products held by a retail savings bank account holder to generate various products