Data org at Torc Robotics (building self-driving autonomous trucks)

Building ETL data workflows across data selection, extraction and human labeling tasks to deliver high-quality datasets for analytics and ML

Designed and Implemented end-to-end data science projects for BESTSELLER (retail fashion) and SONY Pictures (entertainment)

● API-based video metadata extraction pipeline using Python, Flask and Tensorflow to automate and improve character, scene and emotion analysis of TV show content.

● Custom Entity Recognition NLP pipeline using NLTK and spaCy to assist targeted advertising campaigns, by identifying company and brand names from news articles achieving 94% entity-level precision.

● Document processing pipeline to localize and generate table structure using the TableNet model, identify text using pytesseract, and digitize tables from over 50,000 noisy and irregular invoice PDFs.

● Deployed an AI-infused Product Lifecycle Management tool on IBM Cloud for 3 brands.

● Developed similar listings feature using Elasticsearch filters resulting in improved listing visibility

● Created optimized RESTful APIs for migrating Invoicing service to SpringBoot application

● Integrated OpenStreetMap into the listing page; reducing overhead costs by 10%

● Coordinated in a 5-person team for general bug-fixing of backend code in PHP and Laravel

● Automated dashboard editing in Apache Superset, reducing client workload by 10 hours weekly

● Created a demand forecasting model for a fashion website using Deep Learning and Forecasting models 4.5% sMAPE

● Evaluated machine learning algorithms for product recommendation and rule-based algorithms for Feature Selection