Jan 2024 – Present · 2 yrs 2 mos

• Codemod.com builds tools that enable developers to automate their crucial yet undifferentiated tasks, allowing them to ship innovative digital experiences, even within massive codebases largely created by AI.
• I advise the startup on their AI efforts

Jan 2024 – Present · 2 yrs 2 mos

• AppAxon is leveraging Generative AI to solve unique problems in the CyberSecurity Space

Jan 2024 – Present · 2 yrs 2 mos

Nov 2023 – Present · 2 yrs 4 mos

Mar 2023 – Present · 3 yrs

• Nelumbium Capital is a new-age asset management company, with the basic premise that "shocks to the system are not exogenous". We use "Complexity Economics" to understand and analyze systemic risks developing in the financial system. The AI engine forms a critical part of the company's modeling for the risk engine. I advise this wonderful startup on their AI efforts.

Jan 2018 – Jan 2019 · 1 yr · Greater Seattle Area · Hybrid

• Built a team from scratch to over 40 scientists and engineers.
• Evangalized adoption of AI & Machine Learning by proposing innovative applications in Client's operations, massively impacting client's productivity.
• Application of ML technology for diverse domains including Telecom industry, Assembly Line monitoring, Mineral Exploration, Digital OOH Marketing, Finance and Health care.
• SON (Self Optimizing Network): As the demand for mobile bandwith and connections increasing in an exponential manner, the Telecom Mobile Operators need to exponentially scale their operations to provide reliable services to customers. Our client provides this service to leading telecom operators like AT&T. This scaling is currently manual, requires niche technical expertise, extremely expensive and has the biggest impact on the OpEx and CapEx of the operators. SON is a system that maximizes the efficiency of the the mobile networks catering to the users demand dynamically. We built a machine learning based system that predicts the upcoming usage pattern and makes adjustments to the network automatically catering to the dynamic and growing demand, thus drastically reducing both CapEx and OpEx.
• Assembly Line Monitoring: We utilized Computer Vision techniques (Yolo, tracking) to monitor the assembly line for quality monitoring, adherence to safety practices, helping with data driven discovery of productivity issues.
• Political Sentiment Analysis in Twitter Feed: Our client, a specialized solutions provider in psephology, wanted to analyze the political sentiment using the people network on twitter and their political sentiment. We crawled and built a network of individuals and their political leaning and mapped it to the location, providing a valuable input for making psephological predictions
• Mineral Exploration: Pattern recognition in data provided by Bhukosh to identify presence of useful minerals across various survey blocks.

Jan 2009 – Jan 2023 · 14 yrs

• Copilot in Microsoft 365 (M365 Chat): A copilot that helps Enterprise users to retrieve information from across their Enterprise Files, Personal Files, Emails and Chats. Understanding deeply how LLMs behave, how LLMs can be managed, Retrieval Augmented Generation (RAG), Designing Metrics and Privacy and Responsible AI.
• Responsible AI: Built a system to ensure offensive terms and intents are detected and filtered off. Also, ensuring utmost importance to privacy and compliance.
• Personalization in Search: Leveraging Knowledge graphs and Embeddings generated that represent the Knowledge graphs to hyper-personalize the search experience for a user.
• Cortana: NLP to understand user's intent, entity extraction. The Orchestration Engine behind Cortana that works with LU (Language Understanding) models and executes a task. This Orchestration Engine has now evolved into the Orchestration Engine that is powering BizChat.
• Recommendations & Refiners in M365 Search (Patent Awarded): Recommendation of Refiner generation to help users find their email efficiently.

Jan 2009 – Jan 2018 · 9 yrs

• Bing Search: Sort Rank (Patent Awarded): Sort Rank is Bing's variation of Google's Page Rank. Developed Algorithm to determine the relative importance of every webpage in the entire web of documents in Bing Index, based on the popularity of the web page.
• This algorithm had an immense impact on both helping serve relevant search results and saving hundreds of millions of dollars in optimizations in serving Bing Web Search.
• Bing Search: I was responsible for Names rankers (Web Search Relevance Ranking for any query that contains a name). Based on the Bing Index (i.e. nearly majority of the important web pages), we built a knowledge graph about every person whose presence is anywhere on the open internet.
• Web Search queries were parsed to extract the person entity and the information that was being requested.
• This knowledge graph was utilized to rank search results related to the person, for disambiguation and for powering related queries.
• Azure ML: Developed a Experiment Workflow management system from scratch, taking it from conception to shipping. This product has now matured into Azure Machine Learning offered as part of Azure Services. This experimentation system is a cloud based system where thousands of scientists can submit workflows of their experiments, processing Tera and Peta bytes of data efficiently, for caching and sharing experimentation steps with their team, and for training their machine learning models that are ready to be deployed into Production.

Jan 2007 – Jan 2008 · 1 yr

• Worked as engineer in the Platform team. Responsible for performance tuning many widely used tools by developing a scalable design architecture of the same.

Jun 2006 – Sep 2006 · 3 mos

• Developing High Performance Matlab. (Parallel Computing in Matlab)

Jul 2004 – Dec 2007 · 3 yrs 5 mos

• Research Intern working on High Productivity languages in Parallel Computing.

Sep 2003 – Jul 2004 · 10 mos · Columbus, Ohio Metropolitan Area

May 2002 – Jun 2002 · 1 mo

• Developed a Bug Tracking software.