Santra Johny

Built and deployed 3 production-grade XGBoost/LightGBM models to predict customer risk-stage transitions across a multi-class bucket system. Built and deployed a standalone end-to end ML model predicting payment defaults before they occur, achieving 89-91% accuracy. Designed and deployed a Python-based rule engine to classify at-risk customers into 4 risk tiers. Operationalized the pre-delinquency model scores end-to-end through a vendor platform. Partnered with business heads and field teams to translate ML insights into actionable collection strategies; mentored junior analyst on Python workflows.

Trained and fine-tuned multiple machine learning algorithms, conducted hyperparameter optimization, and performed comparative analysis. Developed a responsive chatbot reducing manual interaction by almost 80% with a session-managed Django backend leveraging OpenAI's GPT-o API. Executed data analysis tasks including preprocessing, feature engineering, and implementation of machine learning models and NLP pipelines. Collaborated on research projects involving generative AI and statistical modeling aimed at reducing hospital readmission rates.