As a researcher at Google DeepMind, I study how to make large foundation models efficient & effective.  This includes reducing model size and training & inference costs, as well as support for controllable tradeoffs such as cost vs quality/task complexity.

In previous work (at Google Research), I worked on practical challenges in the design of machine learning systems -- robustness, concept drift, human-AI interaction, user behavior modeling, etc.  I worked  on a number of projects with students and faculty at IIT Bombay, IISc, IIT Madras, IIIT-H, among others.

Research history

I've recently collaborated with several excellent faculty members and their students across a number of institutions, including (in order of recency):

Prior to Google, I managed applied scientist teams that built & deployed large-scale user behavior models for Microsoft Bing Ads -- click & conversion prediction, bidding agents, user preferences, query categorization, etc.

In the distant past, I've developed computational models & theories for neuroscience, cognitive psychology, psychiatry and neuropharmacology.  In  my graduate career, I worked on a variety of brain-computer interface problems, examining the interplay between brain signal decoding and application context.  Even earlier than that, I worked on a number of topics in database systems research. See my publication list for more details.

Academic research groups I've worked at:

Academic history

I earned a Ph.D in computer science at the University of Washington in the Neural Systems Laboratory with a dissertation titled "Brain-Computer Interfaces for Control and Computation".  I received a Bachelor's degree in Computer Science from the Indian Institute of Technology, Bombay. I worked as a post-doctoral researcher at UW and in the Cognitive Science department at the University of California, San Diego. 

I worked on Human-aided Computing (MIT tech review article) in collaboration with Desney Tan at Microsoft Research, and on the Brain-Controlled Humanoid Robot (Discover magazine article) at UW, with Rajesh Rao.

Teaching

In Spring 2021, I presented guest lectures on reinforcement learning in the brain for Prof. Arjun Ramakrishnan's course on neuroeconomics at IIT Kanpur (BSE662A). 

I taught the winter 2012 course on Natural Computation at UCSD, covering a range of introductory machine learning concepts and their applications in theoretical models for cognition & neuroscience. 

I co-managed the Spring 2006 course on Brain-Computer Interfaces at the University of Washington.