I led the development of an innovative product aimed at assessing customer creditworthiness through advanced NLP and machine learning techniques applied to transactional data. This initiative involved the design of sophisticated machine learning models that effectively categorized transactions into micro and macro levels. By leveraging this categorized transaction data and adhering to traditional risk modeling processes, we generated a comprehensive transactional score for customers, enhancing… Show more I led the development of an innovative product aimed at assessing customer creditworthiness through advanced NLP and machine learning techniques applied to transactional data. This initiative involved the design of sophisticated machine learning models that effectively categorized transactions into micro and macro levels. By leveraging this categorized transaction data and adhering to traditional risk modeling processes, we generated a comprehensive transactional score for customers, enhancing our understanding of their financial stability. Beyond my technical contributions, I played a crucial role in the product's conceptualization and the development of tailored use cases for a variety of industries. Furthermore, I successfully led the execution of Proof of Concepts for a diverse clientele, spanning sectors such as insurance and Non-Banking Financial Companies (NBFCs). Show less

Document Classification and Information ExtractionDeveloped a mobile-net CNN model using transfer learning for accurate classification of documents like Aadhaar and PAN cards . Utilized SSD for object detection and Py-tesseract for text extraction. Contributed to project integration and deployment for Strategy one team.Collection ScorecardEstablished population selection criteria for optimal scorecard development. Segmented the population into three scorecards based on… Show more Document Classification and Information ExtractionDeveloped a mobile-net CNN model using transfer learning for accurate classification of documents like Aadhaar and PAN cards . Utilized SSD for object detection and Py-tesseract for text extraction. Contributed to project integration and deployment for Strategy one team.Collection ScorecardEstablished population selection criteria for optimal scorecard development. Segmented the population into three scorecards based on delinquency. Employed advanced statistical techniques for one scorecard. Classified customers into high-risk and low-risk groups. Delivered a valuable risk management tool to the client. Show less