𝗢𝗻𝗹𝗶𝗻𝗲 𝘁𝗿𝗶𝗴𝗴𝗲𝗿𝗶𝗻𝗴 𝗺𝗼𝗱𝗲𝗹 𝗳𝗼𝗿 𝘁𝗲𝗰𝗵𝗻𝗶𝗰𝗮𝗹 𝗾𝘂𝗲𝗿𝗶𝗲𝘀: Built a query understanding model to identify action/intent in a text and related synonyms that can occur in similar context. Applied deep learning techniques like BERT, intent matching to compute query-query similarity. Also, worked on clustering technique to improve the recall further.𝗔𝗣𝗜 𝗹𝗼𝗼𝗸𝘂𝗽 𝗮𝗻𝘀𝘄𝗲𝗿 𝗶𝗻 𝗕𝗶𝗻𝗴: Implemented end-to-end solution to provide developers with quick… Show more 𝗢𝗻𝗹𝗶𝗻𝗲 𝘁𝗿𝗶𝗴𝗴𝗲𝗿𝗶𝗻𝗴 𝗺𝗼𝗱𝗲𝗹 𝗳𝗼𝗿 𝘁𝗲𝗰𝗵𝗻𝗶𝗰𝗮𝗹 𝗾𝘂𝗲𝗿𝗶𝗲𝘀: Built a query understanding model to identify action/intent in a text and related synonyms that can occur in similar context. Applied deep learning techniques like BERT, intent matching to compute query-query similarity. Also, worked on clustering technique to improve the recall further.𝗔𝗣𝗜 𝗹𝗼𝗼𝗸𝘂𝗽 𝗮𝗻𝘀𝘄𝗲𝗿 𝗶𝗻 𝗕𝗶𝗻𝗴: Implemented end-to-end solution to provide developers with quick answers and help on Bing which improved user metrics. Single-handedly owned customer requirement discussions, design specs, UI handshaking, backend architecture design & implementation and online triggering model. Performed A/B testing, evaluated user metrics and deployed solution to production.𝗧𝗿𝗮𝘃𝗲𝗹 𝗮𝗻𝘀𝘄𝗲𝗿𝘀 𝗶𝗻 𝗕𝗶𝗻𝗴: Applied intelligence, analyzed and interpreted data to find engagement opportunities for Bing in Travel segment. Implemented end-to-end solution for powering up rich user experience for travel query intents like countries, states. Show less

𝗔𝗻𝘀𝘄𝗲𝗿 𝘀𝗻𝗶𝗽𝗽𝗲𝘁𝘀 𝗼𝗻 𝗕𝗶𝗻𝗴 𝗳𝗼𝗿 𝘁𝗲𝗰𝗵 𝗿𝗲𝗹𝗮𝘁𝗲𝗱 𝗵𝗲𝗹𝗽: Researched and presented the analysis of technical queries on Bing to get business insights and in the process built text classifier to classify technical queries from billions of records of raw web data. In the next phase, worked on Question-Answering pipeline which included mining text passages and list passages from unstructured web page data, applying deep learning techniques like DSSM, CDSM, Encoders to… Show more 𝗔𝗻𝘀𝘄𝗲𝗿 𝘀𝗻𝗶𝗽𝗽𝗲𝘁𝘀 𝗼𝗻 𝗕𝗶𝗻𝗴 𝗳𝗼𝗿 𝘁𝗲𝗰𝗵 𝗿𝗲𝗹𝗮𝘁𝗲𝗱 𝗵𝗲𝗹𝗽: Researched and presented the analysis of technical queries on Bing to get business insights and in the process built text classifier to classify technical queries from billions of records of raw web data. In the next phase, worked on Question-Answering pipeline which included mining text passages and list passages from unstructured web page data, applying deep learning techniques like DSSM, CDSM, Encoders to build text similarity and passage ranking models and query intent model to improve the precision. This helped in gaining recall by 12% while maintaining 85% precision. Show less

𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗿 𝗮𝗻𝘀𝘄𝗲𝗿 i𝗻 𝗕𝗶𝗻𝗴: Worked on backend solution, web data mining and triggering model to add developer experience on Bing for different programming languages. Also, analyzed solution performance and user metrics on several pivots and implemented several new features, fixed feature bugs which improved the solution performance and other metrics like user click rate by 20%.

𝗗𝗲𝘁𝗮𝗶𝗹 𝗽𝗮𝗴𝗲 𝗱𝗲𝗯𝘂𝗴𝗴𝗲𝗿: Created the Tomcat Java Servlet application, a user interface to automate the process of onboarding and adding a feature to the debugger scope. The implementation allows users to update the resource files through the UI, which automatically generates the backend code. Changed the existing architecture of the Debugger and migrated from file system to SQL database. Languages used HTML, CSS, JavaScript, Java.