Failure Prediction: Built an XGBoost model with MLflow that could predict part failure several months in advance for on-highway engines

Forecasting framework: Created an automated time series forecasting tool that can read the data, build multiple models and display the results in RShiny along with validation and diagnostics Name Cleaning: Cleaned customer names that were incorrectly recorded by measuring the similarity within the entered names and creating appropriate groups in an unsupervised manner using python

Loyalty Segmentation: Helped create customer segments that were used to understand how purchase behavior evolved over time. Business managers were able to utilize this information to make Promotional strategies and key strategic decisions.Price Elasticity : Created Linear Mixed Models to understand how different customer segments reacted to price changes and promotions. Customers segments could be targeted with low or high promotional offers based on their elasticity value. Other Projects : Campaign Performance, New Product Performance Tracking

Technology Company, June 2015 to July 2016 Service center efficiency measurement: Developed a scalable and a white box framework to measure the efficiency of a service center along with metrics to showcase the impact of improving the current model, identified that 10% of problems solved were handled inefficient manner which led to a wastage of 2 hours per employee per day Improve the Website: Studied the movement of customers through the website of a major technology giant to identity the driver of Customer Satisfaction using Logistic regression which helped prioritize problems for solving Retail Company, August 2016 to April 2017 Cross price elasticity: Build Linear Mixed models to measure the effect of change in price of an items to the sales of an supplement item, this metric helped the team identify the products whose price reduction will contribute the maximum benefit Campaign Targeting: Campaigns are run almost every month and the customers are notified via direct mail. Developed a Random forest model to identify the customers to target for an upcoming campaigns based on their response rate and the incremental sales they bring. This helped maximize the Return on Investment (ROI) from the campaign Exception Reporting (Anomaly Detection): Identified changes in key business metrics quickly and more efficiently for numerous metrics using Ensemble model. This framework resulted in saving 15 man hours per week and was able to identify 10% more significant changes in Key Business Metrics Rapid Impact: Quick high impact turn around projects which occur during holiday seasons where multiple campaigns are run in parallel

Worked on a Summer Research Fellowship project on Comparing of Aerosol Optical Thickness data obtained from two different satellites Terra and Aqua.1. Extracted relevant information from open source satellite data2. Aggregated the extracted data through MATLAB3. Created contour plots to visualize the data