Retail Decision Models• Developed an XGBoost model to adjudicate TD Auto Finance (TDAF) Auto loan applicants for the Canadian market utilizing the Bureau, Internal Customer data, and Transactional data.• The Through-the-Door (TTD) XGBoost model outperformed the Traditional Scorecard (Logistic) Model by 4% in KS and AUROC and the Franchised (TD Existing Customers) model outperformed by 7% in K-S and AUROC.• Lift in the model performance led to better discriminate the default risk and translates to ~ 1.5 billion dollars worth more approvals per year. Retail Capital Models• Develop statistical forecasting models to predict the trend of portfolio performance based on customer profiles, product features, and macroeconomic scenarios for various retail asset portfolios, including installment loans and credit cards.• Developed a Probability of Default model using the XGBoost algorithm to classify the defaults for the Other Consumer portfolio (Personal Lending). The Sensitivity ratio of the model forecasts over 13 quarters outperformed the traditional Logistic Regression model by 1.5x.• Designed a proxy for the Unemployment Rate to re-calibrate the Probability of Default model for the TD’s credit card portfolio to mitigate the effect of the sudden surge in UR during the COVID outbreak and attained a 23% reduction in 9Q cumulative Net Charge Offs (NCO).• Interpreted and explained the model forecasting and model monitoring results to the senior management. Performed model performance and robustness testing including Accuracy testing, Stability testing, Sensitivity Analysis, and Scenario Analysis along with model documentation and implementation.Process Automation• Led a team of four members to develop an Automation tool that currently produces the Model Development Report (MDR ~ 300 pages). The automation tool led to an 80% reduction of time invested in producing the report (8 weeks to 60 mins).

• Developed a Probability of Default and Probability of Prepay model using Begg & Gray Individualized Logistic Regression methodology under the hazard rate competing risk model framework for Residential Mortgage portfolio.• The model was used as an input for regulatory stress testing – CCAR, DFAST, MST and allowance calculations under US GAAP/CECL and IFRS-9.• Constructed an exploded panel dataset (landmark framework) using quarterly data from Moody’s Analytics Residential Mortgage-Backed Securities (RMBS) for US Mortgage Market and Internal Residential Mortgage data to develop the model. • Performed variable reduction using Variable Clustering (Principle Component) and Information Value statistic. Weight of Evidence (WOE) was used to convert continuous risk drivers into a set of bins based on the dependent variable distribution. • Back-tested the model on the Internal Residential Mortgage portfolio and external RMBS data using various performance metrics – MAPE (< 7%) and WMAPE (< 5%) across all time horizons - 4Q, 9Q, 13Q and 20Q. • Created Portfolio Sensitivity Analysis Report for the Retail Credit Card (RCS) portfolio and conducted the Annual Model Reviews (AMRs) for Small Business & RCS portfolios. • Conducted scenario analysis using CCAR supervisory stress scenarios to a test a range of macroeconomic conditions including Severely adverse (SSAS), adverse (SAS) and baseline (SBS) scenarios.

Acquired a new client to implement an Integrated Dashboard which contains a holistic view of Business Key Performance Indicators.Understand and translate statistics to address client business problems and design appropriate solutionsDevelop sophisticated analytical frameworks to generate Business Insights and effectively provide consultancy to clients on various aspects of their business using data analytics

Worked on a project titled Optimization of Tangential flow Filtration of nanoparticles.Optimized key parameters of the process such as transmembrane pressure, crossflow rate, membrane area and filtration control.Carried out experiments to determine the dependence of transmembrane pressure on other variables which provided the data for the final optimization of the whole process for the required drug

Digital Marketing( Social Media Marketing)Involved in a 4 member team for the marketing and rebranding of the launch of Explara.com (earlierknown as Ayojak.com), online event ticketing & registration platform offering event discovery to users and event solutions to event organizers.

Managed all the financial transactions of 3 major student managed festivals and as well as small scale campus festivals. Dealt with planning and budgeting of overall transactions of working fund, which constituted more than 10 million INR.

Project:Laboratory Scale Experimental studies to Decolorize Yellow colored machine backwaterInvolved in a 3 member team to find methods using which the yellow colored machine backwaters could be decolorized.It could not be decolorized using the normal techniques. Since the samples contained high amounts of pulp, using flocculants proved to decolorizing the machine backwaters.