Sep 2015 – Nov 2025 · 10 yrs 2 mos · Los Gatos, CA · Hybrid

• I am leading an applied research team focused on Foundation Models as well as Machine Learning for conversational discovery experiences for Netflix subscribers. As of July 2024, I am also building a team to work on AI Inference Optimization Research.
• We research and develop large-scale foundation models, including LLMs, based on structured and unstructured multi-modal data. These are used for various personalization and discovery use cases.
• Interactive discovery experiences include search and contextual recommendations where subscribers choose to express their entertainment needs explicitly in some ways. To optimize the experiences my team works through the full cycle of opportunity understanding, problem formulation, machine learning research, experiments, engineering, and releases.
• In this role, I enjoy leading a stunning team to shape product, research, and engineering vision working very closely with several cross-functional partners.
• Few years ago, in my initial role as a tech lead at Netflix, I identified many opportunities in previously untapped aspects of Netflix Search and rallied the teams to launch product and algorithms enhancements that led to measurable gains in user satisfaction. Some examples include machine learned Instant Search ranking, entity based recommendations within Search, implicit spell correction, optimization of results presentation etc.
• Netflix's Search (and in general Search on streaming media platforms) have some unique challenges compared to the Web Search. Read about those in our paper linked below.

Mar 2007 – Sep 2015 · 8 yrs 6 mos · Sunnyvale, California

• Led a team of research engineers to develop a Contextual Bandit solution for the problem of integrating results from various Yahoo! content verticals (News, Finance, Local, Movies etc.) into the organic Web Search results and optimally triggering and displaying them based on the query intent. Helped architect federated search middleware to serve the models for hundreds of millions of queries on Yahoo! Web Search. This large cross functional effort produced significant metrics wins (CTRs, dwell times, downstream revenue), publications, patents and spawned several associated research efforts.
• Led a team of engineers and applied researchers to develop and deploy a brand new system for Named Entity Resolution and Intent Detection.
• Improved the relevance of machine learned ranking functions and achieving better relevance and revenue trade-off for a $$$ millions Yahoo! product. Developed features, trained models, led A/B tests, performed rigorous model evaluations with editorial inputs, researched new methods with other researchers that led to patents and publications.
• Developed a machine learned ranking approach for selecting the most relevant, most readable, most scannable and most presentable title for a given URL in the Web Search results for a given query.
• Led applied research efforts to build a machine learned ranking solution to improve targeting for Yahoo!'s mobile advertising.
• Led initiatives to better understand user behavior on Yahoo! finance and Yahoo!'s smart TV offering.
• Fields: Web Search, Information Retrieval, Machine Learning, Software Engineering
• Programming: Python, R, Java, C/C++, Shell Script on GNU/Linux platform. PIG on Hadoop.

May 2005 – Mar 2007 · 1 yr 10 mos

• Researched and developed algorithms and software systems for Become’s proprietary crawler-based web search engine technology and its online shopping web site.
• Significant contributions in design and development of a cutting-edge crawl platform that allows continuous crawling of billions of webpages and anytime index updates with efficient resource utilization.
• Researched, designed and developed large scale machine learning applications to improve relevancy and help Search Engine Marketing.
• Fields: Web Search, Information Retrieval, Machine Learning, Software Engineering
• C/C++, Java, Perl on GNU/Linux platform; Octave, Protege, Lucene, Ant, Log4J etc.

Dec 2003 – May 2005 · 1 yr 5 mos

• Worked with Prof. Dr. Rohini Srihari [http://www.cedar.buffalo.edu/~rohini/] on a text mining project named "Unapparent Information Revelation Creation, Visualization and Mining of Concept Chain Graphs".
• Researched on a novel document representation model called Concept Chain Graph and built a prototype system based on it to prove the versatility, richness and suitability of the representation for complex text mining applications.
• Developed a procedure for automatic discovery and ranking of Concept Chains as well as their evidence using probabilistic models and graph search.
• Publications:
• (1) R. K. Srihari, S. Lamkhede, A. Bhasin, W. Dai. "Contextual Information Retrieval using Concept Chain Graphs". Proceedings of the Workshop on Context-based Information Retrieval (Context 05) Paris, France, July 2005.
• (2) R. K. Srihari, S. Lamkhede, A. Bhasin. "Unapparent information revelation: a concept chain graph approach". CIKM-2005 [Poster]

Dec 2002 – Jul 2003 · 7 mos

• System understanding, module development and technology leveraging.
• Worked on two portal development projects for claims intake for an international insurance firm, MetLife.
• Gained invaluable experience in Web Services, J2EE application development and CMM level 5 process.