Sharmita Dey

Sharmita Dey completed her Master’s degree at the Institute of Artificial Intelligence, TU Dresden, Germany. She conducted her master’s thesis on "Generalized Decoding of Control Signals from Surface Electromyography Signals" at the DLR Institute of Robotics and Mechatronics in Oberpfaffenhofen, Germany.

Following this, she earned her Ph.D. degree from the Department of Computer Science, University of Göttingen, Germany, focusing on "Learning-Based Biomimetic Strategies for Developing Control Schemes from Lower Extremity Rehabilitation Robotic Devices," for which she was awarded summa cum laude. During her Ph.D., she interned at NASA's Jet Propulsion Laboratory, contributing to projects enhancing ground robot traversability in challenging environments. In this role, she became a part of team CoSTAR taking part in the DARPA Subterranean Robotics Challenge 2021.

She then held a postdoctoral position at the University of Göttingen, Germany, where her research focused on vision-based, label-efficient learning and object-centric models for understanding the interactions and dynamics of moving objects.Further, as a postdoc at the University of Göttingen, her research explored domain adaptation, multitask learning, multimodal learning, and learning from simulated interactions using world models, to enhance predictive and adaptive learning.

Currently, at ETH Zurich, her research will focus on generative design of assistive aids and multimodal self-supervised learning. Alongside her research, she is an active mentor in Women in Machine Learning (WiML) and Machine Learning for Health (ML4H), programs that support women in machine learning. She also serves as a reviewer for leading AI and robotics venues, including NeurIPS, ICLR, ICML, AISTATS, ICRA, and IROS.