Created a platform for Declarative Config based extraction of E-commerce verticals. Keys characteristics of the system1) Easy to create Config for doing shallow or deep extraction. 2) Should support historical extraction i.e. System should able to extract information given the historical webpages. 3) Ease of debuggability in case of in-correct/in-complete data is getting extractedAutomated Layout Element Detection from E-commerce Pages1) For doing deep extraction1, identification of Layout element was automated. This helped us in automatically identifying such element, and benefited us in multiple scenarios mentioned below. These findings also got published in NIPS 2015 workshop. - Less manual effort for robot developer - Generating report for new types of LayoutsWhile implementing this, multiple technology are used from Selenium, Python, Classification algorithms

Information RetrievalScalable Crawling System

Machine Learning Algorithms on Cloud:- We implemented various Machine Learning algorithms using Divide and Conquer Approach in a distributed environment using Hadoop. Basic concept was to divide the data into multiple chunks of smaller size, do learning on each of them and combine it using different methods such as majority.