Rutav Shah

I am a first year Computer Science PhD student at UT Austin.

The goal of my research is how robotic agents can acquire generalized behavior robust to domain and task variability. For my undergraduate thesis, I was co-advised by Prof. Abir Das (Computer Vision and Intelligence Research Lab at IIT Kharagpur) and Dr. Vikash Kumar (FAIR Pittsburgh).

Apart from research, I love teaching, trekking, and running! Some high altitude treks I have completed, Tapovan, Brahmatal in the Garhwal region and Buran Pass in the Himachal Himalayas.

Be kind, cheers!

Email  /  CV  /  Google Scholar  /  Twitter  /  Github

profile photo

My research interests lie broadly in Robotics and Machine Learning. I am interested in building autonomous agents that can acquire complex behaviors in unstructured, un-instrumented settings like home, hospital, etc., with little to no human intervention. Currently, my research focuses on how generalisation of an agent can be improved via learning "better" representations.

Contrast and Mix: Temporal Contrastive Video Domain Adaptation with Background Mixing
Aadarsh Sahoo*, Rutav Shah*, Rameswar Panda, Kate Saenko, Abir Das
35th Conference on Neural Information Processing Systems (NeurIPS), 2021.
project page / arXiv

Novel temporal contrastive learning approach for unsupervised video domain adaptation.

RRL: Resnet as representation for Reinforcement Learning
Rutav Shah, Vikash Kumar
International Conference on Machine Learning (ICML), 2021.
project page / video / arXiv

Standard image classification models used in conjunction with standard IL or RL pipelines can efficiently acquire behaviors directly from proprioceptive inputs.

Webpage template courtesy : Jon Barron