• Delivered ~$25 million annual savings by reducing ~2700 man hours weekly by automating property valuation.Built an Automated Valuation Model (AVM) using TensorFlow-based custom nearest-neighbor architecture to identify comparable properties to predict prices for 4 major property types – Multi-family, Office, Industrial, Retail.• Established an end-to-end data pipeline, & refactored complete code following PEP guidelines to productionise a deep learning real estate valuation model in the client's environment.• Developed a Revenue forecast model for a major European EV supplier leveraging Time Series forecasting packages such as fb-prophet, neural prophet. Reduced time taken for forecasting from 1 week to 4 hours.• Created a heuristic based driver scheduling algorithm to incorporate driving, terminal, and regulatory constraints to create a simulation of bills movement across 34 terminals across United States for a long-haul trucking client.

• Built Sales Forecasting Model using Keras for creating promotion strategy for an American online clothing chain.• Developed an end-to-end data pipeline on Azure utilizing Azure functions.• Utilized KNN for determining comparable real estate properties, incorporating weights for different features.

• Developed several Keras models for classifying YouTube Thumbnails and established data pipeline for extracting thumbnails using YouTube API.

• Animated & Produced lectures for YouTube & garnered 10,000 hours of watch time.• Managed company website, and social media to promote business.

• Created a web app using Django & Three.js to make 3D human models with specific dimensions, to be used as a virtual mannequin for an online clothing store.