Research Scientist Intern, Meta Reality Labs (May 2022 - Ongoing)

GNN-based physics solver for time-independent PDEs - developed two GNN architectures for solving static physics problems, which outperform the current surrogate model options available and have great generalization capability.

1. CTR and Win Probability Prediction and Fraud detection (Digital AdTech)
Developed a real time online predictive algorithm for Click Through Rates (CTR) and bid win probability prediction for a leading digital media company.   Also, built real-time fraud detection engine to flag fake ad requests from dummy users and bots.

2. Customer Analytics Solution (Retail)
Developed a customer analytics solution to uncover customer purchase patterns, predict repeat purchase rate of customers and purchase value of repeat purchases for a large brick and mortar retailer. Also, designed a Tableau Dashboard to track the performance of the campaigns to understand – 1. Promotions & Customer Engagement, 2. Top problems seen across Brands and attributable factors, 3. Performance of campaigns Week-on-Week. 

3. Category Share Analysis (Automobile)

Identified key enablers responsible for increasing Category Share of Moped sales, for a leading automobile company in India. Built a Mixed Effects model by analyzing data from various sources – Customer Demographics, Retail Finance, Campaign Data, Price and Competition data. Provided recommendations at a state and town class level to improve the moped category share.

4. Price Elasticity Modelling

Built a price recommendation engine for a leading American fast food company at store level granularity. Used regularized regression for modelling the price elasticities of for thousands of items across 14000 stores. Developed a prize optimization framework to use the modelled price elasticity to give the optimum price for every item at a store level maximizing the revenue of the store.