Aug 2015 – Aug 2024 · 9 yrs · Mountain View, California

• (2022 - 2024)
• Senior Staff Research Scientist at Google Research (working on Holistic NLP / LLM Research)
• Lead and manage (TLM) team ("Hummingbird") of engineers and researchers working on different research problems pertaining to LLMs including:
• Improving instruction / prompt-tuning / few-shot capabilities of LLMs via better algorithms and training objectives.
• Improved in-context learning algorithms for LLMs.
• Metric-driven, holistic optimization and finetuning of LLMs.
• Scalable, Retrieval and Knowledge augmentation of LLMs.
• I lead some of the highest impact research efforts impacting and deployed in Google Ads systems.
• (2019 - 2022)
• Staff Research Scientist at Google Research (working on Scalable, Multilingual, Multimodal NLP)
• Led and managed (TLM) team ("Omniglot") working on cutting-edge, scalable NLP research pertaining to Representation Learning, Retrieval and Relevance modeling of multilingual, multimodal objects.
• Our work was deployed in some of the highest-impact systems -- especially in the eCommerce domain at Google.
• (2017 - 2019)
• Senior Research Scientist at Google Research (working on Multilingual NLP)
• Started and led new research team ("Omniglot") working on cross-lingual transfer learning and multilingual representation learning.
• Led to dozens of the most widely used Google Search and Ads product / features being launched in new countries across the globe using a novel research algorithm I designed and productionized.
• Served as lead for a larger cross-team NLX i18n group.
• (2015 - 2017)
• Research Scientist at Google Research (working on Health NLP)
• Worked on different NLP understanding research problems in the health / medical domain.
• Started and led critical life-saving work on self-harm prevention using powerful machine learning modeling. Represented Google at different forums on this topic such as the 2017 Pentagon summit on suicide prevention.

May 2014 – Aug 2014 · 3 mos · Mountain View, CA

• Developed automatic knowledge extraction framework as part of the Knowledge Vault team.

May 2012 – Aug 2012 · 3 mos · Redmond, WA

• Worked on understanding the role of intrinsic diversity in web search as part of the CLUES (Context, Learning, and User Experience for Search) team.

May 2011 – Aug 2011 · 3 mos · Redmond, WA

• As part of the Machine Learning and Intelligence group, I developed the correctable learning framework and algorithms for this learning model.

Aug 2010 – Jun 2015 · 4 yrs 10 mos · Ithaca, New York Area

• Responsible for maintaining the experimental text-search engine at arxiv.org. Also develop recommendation algorithms for my.arxiv.org. Launched and maintain peer-grading service at peergrading.org.

May 2009 – Aug 2009 · 3 mos · Bengaluru Area, India

• Researched the use of non-negative matrix factorization and topic models within a pseudo-relevance feedback framework.