Dr. Amartya Sanyal
Dr. Amartya Sanyal
Lecturer at the Department of Computer Science
Additional information
Research area
I am an ETH AI Center Postdoctoral fellow where I work closely with Prof. Fanny Yang and Prof. Bernhard Schölkopf. My main research interests lie in notions of reliability including privacy, adversarial robustness, distribution shift, fairness, and calibration and whether modern machine learning algorithms fail at them. Some examples of questions that I am currently interested in are a) whether there are statistical and computational trade-offs between these notions for modern machine learning algorithms, b) what is the role of proper representation learning in achieving these notions tractably, and c) how to design algorithms to implement reliable machine learning in practice.
I finished my D.Phil (PhD) at the Department of Computer Science in the University of Oxford in 2021, where I was advised by Prof. Varun Kanade and Prof. Philip H.S. Torr. My thesis was titled Identifying and exploiting structures for reliable deep learning. During my PhD, I also had the opportunity to spend time at Facebook AI Research with Dr. Edward Grefenstette and the Laboratory for Computational and Statistical Learning in Genoa with Prof. Lorenzo Rosasco.
Prior to that, I did my undergraduate (2013-2017) at the Indian Institute of Technology, Kanpur where I worked closely with Prof. Purushottam Kar and Dr. Prateek Jain.