Jun 2022 – Present · 3 yrs 11 mos

Oct 2021 – Jan 2023 · 1 yr 3 mos · San Francisco Bay Area

• Led a team of applied scientists and engineers to build real-time agent assistance for chat and voice, which includes features such as conversational search, knowledge extraction, recommendation, suggested responses, and completions using the latest AI/ML technologies such as semantic matching, neural ranking, and Large Language Models (LLMs).
• Driving overall AI vision by being part of the leadership team.
• Collaborated with cross-functional teams of sales, success, marketing executives, and enterprise customers.

Oct 2019 – Mar 2021 · 1 yr 5 mos · San Francisco Bay Area

• Conversational AI, Entity Linking, and Resolution, Q&A, Social Bots

Aug 2018 – Sep 2019 · 1 yr 1 mo · San Francisco Bay Area

• Speaker identification, ASR etc.

Mar 2013 – Aug 2018 · 5 yrs 5 mos · San Francisco Bay Area

• Worked with Query understanding, search verticals (Maps, Web, App etc), search results rendering, knowledge graph, and ranking teams to improve search for Spotlight, Safari, Siri.
• Led a team to build maps instant search and autocomplete system (Infrastructure, indexing, retrieval, query building, spelling correction, ranking and client-server interaction) from scratch to improve query conversion for all of the top locales (including Japan/Korea) significantly which drives major part of search conversion, shortens session length, reformulations and abandonments significantly, improves ranking metrics and meets an optimistic latency SLA at very high throughput to support maps app, third parties, weather app, calendar app, safari, spotlight, Siri etc
• Built a Geo knowledge graph using WFST incorporating all the maps data and learned signals from CTR to unify query completion, contextual query understanding and interpretation ranking for autocomplete and search.

Nov 2011 – Mar 2013 · 1 yr 4 mos · San Francisco Bay Area

• Led a team to build a scalable recommendation pipeline utilizing billions of data points for evergreen, trending, a good set of URLs computed several times per hour.
• Developed a scalable, config-driven framework for topic modeling, user-user, and user-item similarity on Hadoop and MapReduce to provide vertical and horizontal scalability.
• Implemented and owned various personalization recommendation algorithms such as like-minded/similar users, similar pages, events driven, location sensitive, trending locally, user favorite/curated list, and keywords-based recommendations.
• Developed rating prediction and evaluation frameworks to evaluate various recommendation methods.
• Collaborated and mentored various teams for various projects and presented work across the org.

Oct 2010 – Nov 2011 · 1 yr 1 mo · Gurgaon, India

• Led a team of several developers, researchers, QA and operation engineers to deliver migration of relational DB to Hbase for analytics, recommendation and various reports generation based on time of day, geography and keyword performance
• Designed and developed a distributed hybrid system (In memory + hbase + disk based) for custom data integration to
• serve complex user queries/reports in real time (milliseconds for complex custom queries involved millions of rows).
• Improved efficiency of reports and recommendation generation for all customers significantly to 10 mins from hours
• Delivered various sessions/workshops on hadoop, hbase, platform architecture, recommendation and scalability.

Aug 2009 – Oct 2010 · 1 yr 2 mos · Berlin Area, Germany

• Delivered lookahead suggestion and spelling correction based on local context, user context and automatic identification of geo context using CTR and maps data
• Built fuzzy search comprising of relevance ranking, local vs global interpretation of query and spelling correction.
• Developed a system for automatic tagging of entities in search queries to maps entities to estimate volume of various query classes, calculate popularity, entities co-occurrence, identify missing geo data, adjusting city popularity (most important city vs. less important city with same name) and to learn most used and known alternates names of entities.
• Led the team for one box search and scrum master

Jun 2008 – Aug 2008 · 2 mos

• Built IBM product recommendation system using spectral clustering and graph cut
• Analyzed financial data using statistical methods.

Sep 2007 – May 2008 · 8 mos

• I worked with Pattern Recognition and Machine learning group for the project based on learning active shape model for damaged plant leaves. It included implementation of code in C++, using OpenCV computer Vision Library and to use some existing MATLAB prototype. I dealt with PCA based segmenting of leaf from background and calculated image based statistics and ASM.

Jan 2006 – Sep 2007 · 1 yr 8 mos

• Worked on Google Technologies (GFS, Mapreduce, Sawzall, Bigtable, workqueue etc) for Google finance (finance.google.com), anomalies detection on time series data and generating training data for News classification system
• Studied News, Crawler, Blog and various other systems for contributing to an experimental project (20%).
• Mentorship and part of interview committee.

Aug 2005 – Jan 2006 · 5 mos