GAURAV PURVA

• Developed a Client Risk Assessment model to identify Money Laundering risk across 15 million SCB retail banking customers worldwide
• Implemented Logistic Regression to identify optimal weights of risk factors that could result in SAR events within the next 90 days
• Engaged bank stakeholders as experts, leading discussions, and integrating their feedback for model calibration
• Engineered features related to CASA transaction activity and client profiles to identify potential risk factors
• Binned features and used weight of evidence as model input
• Evaluated model performance using expert ratings
• Responsible for the management of 3 resources on the model development team, and ensured the timely delivery of the model
• Responsible for model development, getting model validated by Global Model Validation team, collaborating with model implementation team and performing UAT before model go-live
• Promoted from Associate Manager to Associate Director in April 2023

Project 2: Pricing of different products aiming to improve sales margin for a Sweden-based retail client

• Changed pricing strategy from uniform to varying prices across store clusters, resulting in a 3% increase in sales margin
• Implemented K-means algorithm using store demographic variables to cluster stores located in several parts of the country
• Used Log-Log Regression with price, discount, and other variables to identify price elasticity. Daily demand was the dependent feature
• Optimized product pricing and maximized margin using the Gradient Descent Algorithm within client constraints
• Developed a methodology to measure the performance of new prices. Responsible for weekly reporting of impact after price change
• Promoted from Data Science Analyst to Data Science Consultant in May 2021

Project 1: Pricing of different products aiming to improve sales margin for a Sweden-based retail client

• Performed data cleaning and preparation steps followed by implementing the Holt-Winters forecasting method on sales data
• Minimized stock-out situations by implementing inventory planning in the R tool which used the forecasted and historical sales as input

• Created a framework to analyze tweet sentiment and researched public opinion of clients and customer satisfaction
• Implemented preprocessing techniques such as stop words removal, repeated letters normalization, and stemming text
• Implemented Naive Bayes technique to classify sentiment of tweets into three categories. Used Python as programming language