Enriching an eCommerce Catalog using Deep Learning

In order to provide a great user experience for customers browsing hundreds of millions items, a retail catalog needs to be augmented with several pieces of metadata. This includes assigning products to types and shelves, extracting product features such as brand and color, determining if two products from different sellers are duplicates, finding the best content from multiple sources to display on the item page, etc. I will describe how we are solving these problems using deep learning techniques such as Convolutional neural networks for text and images, Long Short Term Memory networks for sequence to sequence modeling, siamese networks and deep autoencoders for duplicate detection.

I am currently a Staff Data Scientist at WalmartLabs and lead the Product Catalog Data Science team. My team works on product matching, product categorization, attribute extraction, content ranking and catalog quality measurement. Prior to this, I worked on large scale product classification at Adchemy to enable Google and Bing ad campaigns for Adchemy's retail clients. I have obtained a Ph.D. and an M.S. degree in Mathematics from University of Michigan, Ann Arbor and a B.Tech. degree in Engineering from Indian Institute of Technology, Bombay, India.