Hello!
I am a resarcher in Machine Learning. I did a PhD at the Department of Computer Science in the University of Oxford, where I was advised by Shimon Whiteson. My long term research goal is to create theoretically grounded methods for intelligent agents which scale to large, real-life scenarios.
My research at Oxford was supported by J.P. Morgan A.I. Fellowship, Google-DeepMind Scholarship and Draper’s Scholarship. Previously, I have also spent time at DeepMind, J.P. Morgan A.I. and Nvidia. Even before that, I did my undergraduate degree in Computer Science at the Indian Institute of Technology, Delhi after which I worked as a research scientist at Xerox Research. I currently work at Amazon where I lead the efforts at the intersection of Reinforcement Learning and Generative AI for text and image domains. My CV can be found here.
Interests
I am interested in understanding the principles of intelligent behaviour and creating algorithms which can help machines learn in practical settings. My current work focuses on themes like Generative AI, large-scale Reinforcement Learning (RL), Natural Language Processing (NLP), Computer Vision, Multi-Agent Systems, Learning Theory, Safety/Alignment in AI with particular emphasis on computational and mathematical foundations. I am also interested in alternate paradigms for learning intelligent behaviour especially towards equipping agents with attributes that help capture aspects like symmetry, compositionality and reasoning under uncertainity in rich multi-modal environments.
News
- Checkout our work on Generalization Across Observation Shifts in Reinforcement Learning, full paper here.
- My PhD thesis Reinforcement Learning in Large State Action spaces is now available online here.
- Checkout our work on Generalization in Cooperative Multi-Agent Systems, full paper here.
- Checkout our work on Open-Ended Learning which leads to Generally Capable Agents, full paper here.
- Our work Tesseract using tensor decompositions for factored action spaces accepted for ICML 2021
- Our work UneEVen using successor features for exploration in multi agent RL accepted for ICML 2021
- I will be starting a research internship at DeepMind in Trinity 2021
- Our work RODE using role decomposition for multi agent RL accepted in ICLR 2021
- Honoured to be awarded IBM PhD fellowship, 2020
- Honoured to be awarded J.P. Morgan A.I. Fellowship, 2020
- I will be starting a research internship at J.P. Morgan A.I. in Michaelmas 2020
- I will be starting a research internship at Nvidia in Michaelmas 2019
- Our work MAVEN using mutual information for efficient exploration in multi agent RL accepted for NeurIPS 2019
- Our work VIREL proposing exact equivalence between RL and inference accepted for NeurIPS 2019
Research Opportunities
I am happy to discuss and collaborate on topics related to my research interests. If you are interested, please send me an email. I especially encourage students from under-represented groups and motivated graduate students to reach out.