•Responsible for the development of business and regulatory models using machine learning techniques to be used for customer management strategies.•Managed the impairment for the card portfolio conducting macro-economic sensitivity analysis and assessing the impact of policy changes on losses.•For any model development task - Collaborated with multiple stakeholders for requirement elicitation, data gathering/preparation/cleansing, model development, validation, implementation, monitoring and post-implementation support.•Performed in-depth analysis around default definition, segmentation, modeling techniques, business impact assessment (ex-swap-set analysis) and answering on-demand business questions. •Showcased to senior stakeholders the material benefit of new models [Collaborative Artifact].

Development of credit risk models for various Banking portfolios.Provide assistance in scorecard implementation in compliance with internal and external guidelines. Performed swap in swap out analysis and provide recommendations to the country risk team.

Built Ensemble model to predict probability of default of a customer in first repayment cycleUsed Feature Selection, Clustering and Recursive Feature Elimination methods for feature selectionExtensively utilized machine learning techniques to develop quantitative models;Built Logistic Regression(L1 &L2), Random Forest and Gradient Boosting modelsSelected models from each technique using k-fold cross validation with KS as performance metric

Developed and utilized the probability of default models for both new and existing customers for sub-prime lending portfolio in UKAnalysed bureau data pull costs' impact on revenues, leading to recommendations which resulted in cost reduction by 20% Ran Test – Control strategies to find new profitable customer segmentsBuilt the master database at client, customer & order level by consolidating and processing data from raw MySQL tables for an online retail finance business in USCART model on ACH clients to find the most predictive variables of bad rate.Conceptualized the process flow for census work involving termination analysis of payroll clients to determine the eligibility