Sarthak Kamat

sartk at berkeley dot edu

I'm a researcher at HRT AI Labs (HAIL).

Previously, I was a research resident at Berkeley AI Research advised by Trevor Darrell, Jitendra Malik, and Ilija Radosavovic. I studied Computer Science at UC Berkeley from 2020-2024.

Research

• Learning Humanoid Locomotion over Challenging Terrain
  Ilija Radosavovic, Sarthak Kamat, Trevor Darrell, Jitendra Malik
  arXiv website

• Humanoid Locomotion as Next Token Prediction
  Ilija Radosavovic, Bike Zhang, Baifeng Shi, Jathushan Rajasegaran,
  Sarthak Kamat, Trevor Darrell, Koushil Sreenath, Jitendra Malik
  arXiv website

• Revisiting Generalizability in Deepfake Detection
  Sarthak Kamat, Shruti Agarwal, Trevor Darrell, Anna Rohrbach
  ICCV workshop 2023 (oral)

• Surface Spherical Encoding and Contrastive Learning for Virtual Bone Shape Aging
  Francesco Caliva*, Sarthak Kamat*, Alejandro Morales Martinez, Sharmila Majumdar, Valentina Pedoia
  MIDL 2021 (special issue)

Projects

• In-Hand Cube Reorientation
  I trained an RL policy for in-hand cube reorientation with the Allegro hand, in the IsaacGym simulator. The model gets RGB pixels in, and outputs motor actions.

• Raytracing Schwarzschild Black Holes
  Aseem Doriwala, Sarthak Kamat, Vincent Lim, Jeremy Ferguson
  CS 184: Computer Graphics, Project Showcase Winner, 2023
  website showcase

• NP-Hard Combinatorial Optimization
  Araav Patel, Sarthak Kamat, Vincent Lim
  CS 170 Fall 2021, 4th place out of 276 teams
  leaderboard writeup

Awards

• Project Showcase Winner, CS 184 Computer Graphics, UC Berkeley - 2023
• 4/276 Place, CS 170 NP-Hard Combinatorial Optimization, UC Berkeley - 2021
• National Merit Scholarship - 2020
• Gold Medal, National French Contest - 2020
• Most Innovative Project, LancerHacks - 2020
• First Place, DV Hacks II - 2020
• First Place, MatadorHacks - 2019

Teaching

• Course Assistant and Reader for EECS 127 (Optimization Models in Engineering), Spring 2022.
• Academic Intern for CS 61A in Fall 2021, the intro CS class at UC Berkeley.
• I taught machine learning to middle schoolers at public libraries in San Jose with Sahil Jain.

Misc

• Africa by Toto on the electric guitar.