Video CoreML and Ranking Foundations (FB Reels)

Working with cross-functional teams (Engineering, PM, UI/UX, Marketing, Field, Finance, Leadership) to build data products and running product led growth campaigns for customer life cycle management for M365 customers and partners.

Worked at the intersection of M365 business marketing and engineering. Built data-driven products to enable smarter recommendations for admins and end-users to ensure the best usage of M365 services. (Think churn, upsell, next best action recommendations, propensity models).Hack4Good(Volunteering) - Worked with Orca Conservancy(non-profit) to build a bioacoustic call recognition and alerting system for Southern resident killer whales (Orcas). This hack project won the Microsoft AI4Good grant, which provided $30K in azure credits for a year.

Leading an agile team of Data scientists and engineers to deliver data solutions which include -- Providing commodity forecast for Non-Fat dry milk spot price in the futures market which results in $700K savings annually by improved forecast (4 cents) and risk management decisions. - Built prediction model for proactively identifying SKUs at risk in any retail store for both our own and competitors product, leading to an annual benefit of $0.5MM - $0.7 MM on proactive corrective actions.- Built a customized bid pricing model for K-12 schools by analyzing past bidding data with the potential of increased profit of $4.5MM.

- Built a machine learning based targeting strategy for finding the best operators(restaurants/chain/hospitals/universities/k-12) to focus out of 1.6 Million operators, leading to increases in conversion rate by 6% and $1.5MM in additional profits- Built Image collection, processing, and prediction pipelines to solve machine vision problems using Deep learning in multiple use cases relating to tracking sustainability practices using satellite imageries and predicting herd animal health scores- Designed AI in E-commerce strategy to build use cases around personalization for various parts of e-commerce websites like user-item recommendations, item-item recommendation, email targeting, cross-sell/up-sell, etc- Running marketing mix models for animal feed categories to understand the effectiveness of various marketing activities

Enhancing corn yields for farmers by improving seed recommendations using a customized growing condition based similarity metrics devised by combining three open source datasets for over 2 million farms using Python and ArcGIS

Assisted Prof. Gedas Adomavicius and Prof. Panagiotis Adamopoulos in an annual recommender system competition to predict through 3.4 billion user-item combinations by ensembles of xgboost and matrix factorization using Python

- Designed supply scheme with Regional head to cut down per trip cost by 50% in 3 months from INR 130 to INR 60- Designed an active cab prediction system to balance demand and supply, reducing cab rejection from 5% to 1%- Developed fraud detection model and automated penalty imposing system saving $300,000 annually using Hive and Python- Worked with Operations on call center optimization and designed task force schedules to keep missed calls within 2%- Automated reporting of customized queries, reducing report generation from 4 hours to 1 hour in R

- Generated reports on mobile application, analyzed user flow and interactions for better understanding of website traffic; delivered findings to Product Manager team to inform future enhancements to app using Adobe Omniture- Predicted conversion rates using R based on various business parameters for better inventory and technical planning

- Analyzed relationships among pharmacies, accounts, physicians at retail and non-retail level by leveraging commercial and internal data sources using SAS and built a dashboard to identify key influencer in the network- Organized 3rd party demand forecasts from 5 different sources; merged, sorted and semantically reconcile unrelated data by category using string matching algorithms for World’s largest retailer- Mentored new hires on SAS language; took seminars on R on topics like Knitr, R shiny, ggplot2 and google viz