Apr 2019 – May 2020 · 1 yr 1 mo · Bengaluru, Karnataka, India

• ORG BUILDING
• Responsible for hiring software engineers by conducting coding, machine learning and System Design interviews. Conducted over 350 interviews over 5 years.
• Grew Technical team from 4 software engineers to 40 software engineer & Data Scientist by hiring, onboarding and mentoring new members over 5 years.
• E-COMMERCE SEARCH PAIN PREDICTION
• Companies employ hundreds of human labelers to evaluate the performance of Search Engine. Problem statement is to replace human effort with automation powered by Machine learning.
• I built a Machine Learning System that predicts the user satisfaction of a Search Result Page by modeling user behavior features. It resulted in F1 of 72 and reduced the need of labelers by 60% saving 1.2 million annually for the company.
• E-COMMERCE SEARCH QUERY CORRECTION
• Standard spell correction approaches do not work well for the new generation of Internet users from Rural India who are not proficient in English.
• I designed context aware character level deep learning model for spell correction that tackles the misspelling originating from regional pronunciation of English words. My system led to 28% accuracy improvement on a curated set of misspellings thereby increasing accurate search queries. It was pivotal to the success of voice search launch by Flipkart

Jun 2017 – Mar 2019 · 1 yr 9 mos · Bengaluru, Karnataka, India

• E-COMMERCE SEARCH QUERY INTENT PREDICTION, TAIL QUERIES
• Query Intent - a sequence labeling problem - is pivotal for retrieving relevant results. However, companies don’t have enough click-log data for tail queries, leading to poor Performance.
• I collaborated with a group of Research Scientists and developed a generative model to improve the performance for tail queries
• My research was later accepted as a Full Paper at the ECIR 2020 Conference.
• E-COMMERCE SEARCH QUERY CLASSIFICATION
• Search Query Classification - mapping search query to taxonomy categories - is pivotal for performance of Search Engine.
• Designed and implemented Deep Learning Based NLP model that led to 11 bps increase in Search CTR that translates to $76 million increase in GMV.

Jan 2016 – Jun 2017 · 1 yr 5 mos · Bengaluru, Karnataka, India

• ECOMMERCE SEARCH QUERY INTENT PREDICTION
• Query Intent - a sequence labeling problem - is pivotal for retrieving relevant results.
• Designed and implemented CRT and LSTM based models that led to 14 bps increase.

Jan 2015 – Jan 2016 · 1 yr · Bengaluru, Karnataka, India

• ATOMING UPDATES TO SOLR DOCUMENT
• Back in the day, SOLR didn’t support atomic updates out of the box.
• Built a storm-kafka based streaming system to update SOLR index with atomic updates that Reduced index building time by 350%.

Apr 2013 – Dec 2014 · 1 yr 8 mos · Ahmedabad Area, India

• ECOMMERCE SEARCH ENGINE FOR BOOKS
• Designed and implemented an Apache Solr based system that Improved update Datastore latency by 180%.
• Designed Machine Learning Algorithm that resulted in an 2.3 % increase in Search Click Through Ratio.
• WORKFLOW ENGINE
• Large scale e-commerce companies have numerous business processes. Workflow Engine manages these processes. In particular,it facilitates the flow of information, tasks, and events.
• Designed a state-machine framework where one defines states and transitions.
• It was initially launched for Customer Service processes. However, owing to simplicity and scalability, it ended up being the backbone of every business process in the company. It was also a key in scaling the company's initiative to launch buildabazaar.

Dec 2012 – Mar 2013 · 3 mos · Ahmedabad Area, India

May 2011 – Mar 2013 · 1 yr 10 mos · Ahmedabad Area, India

• TECH LEADERSHIP AND MARKETING
• Built the initial prototype of the product and led the team of 7 engineers to build the product from scratch.
• Shipped product to 153 sites with bi-annual subscription.