Apr 2023 – Dec 2024 · 1 yr 8 mos · Kolkata, West Bengal, India · On-site

• Developed and Validated PD Model: Created a Probability of Default (PD) model using logistic regression, achieving a Gini coefficient of 40.43%, 85% accuracy, and an AUC-ROC score of 0.75, and monitored its consistency with the Population Stability Index.
• Refined Risk Assessments: Calculated Loss Given Default (LGD) and Exposure at Default (EAD) for 43,000 defaulted accounts, enhancing the precision of loan portfolio risk assessments, and calculated Expected Loss (EL).
• Improved Model Performance: Implemented outlier detection, missing value imputation, Weight of Evidence (WOE), and Information Value (IV) to reduce model complexity by 30% and improve performance by 7.3%.
• Optimized Credit Scorecards: Collaborated on optimizing scorecards by calculating parameters and fine-tuning thresholds to balance loss reduction with customer coverage.
• Market Mix Modelling for Revenue Increase: Developed market mix models for an ecommerce platform's product subcategories, using linear regression techniques (additive, multiplicative, Koyck, distributive lag) to optimize budget allocation, resulting in a 15% revenue increase.
• Comprehensive Data Analysis and Model Validation: Analyzed over 1.6 million rows of data using feature engineering and achieving an R² value of 0.82; validated models through cross-validation, checking for multicollinearity and heteroscedasticity.
• HR Dashboard Implementation: Pioneered the development of a live HR dashboard, boosting data accuracy by 20% and reducing reporting time by 30%.

Nov 2022 – Dec 2022 · 1 mo

• Collaborated on an energy analysis project, processing data from 768 building shapes to predict energy consumption, rectifying 528 null cells, and analyzing 1296 samples.
• Developed precise predictive models for Heating and Cooling Load, achieving high accuracy with advanced regression techniques like Decision Trees and Gradient Boosting. Introduced key features like "Overall Load" and "Efficiency" for improved energy management.
• Contributed to impactful research in machine learning and data analysis, enhancing building energy management efficiency by 15% within a team setting