Harsh Satija

I'm a postdoctoral fellow at the Vector Institute, collaborating primarily with Gillian Hadfield. My research interests encompass the AI alignment, collective-intelligence, infrastructure and collaboration mechanisms for agents, sequential decision-making under uncertainty, and reasoning in Large Language Models. I obtained my PhD from McGill University in collaboration with Mila, under the supervision of Joelle Pineau, where I focused on alignment challenges in various Reinforcement Learning settings. My doctoral work has been featured in leading machine learning conferences. Previously, I also spent some time as a software engineer in the industry.

• Group Fairness in Reinforcement Learning

  Harsh Satija, Matteo Pirotta, Alessandro Lazaric, Joelle Pineau

  TMLR 2023

  Oral at EWRL 2022

  Paper

  Github

• Multi-Objective SPIBB

  Harsh Satija, Philip S. Thomas, Joelle Pineau, Romain Laroche

  NeurIPS 2021

  Paper

  Github

• Constrained Markov Decision Processes via Backward Value Functions.

  Harsh Satija, Philip Amortila, Joelle Pineau

  ICML 2020

  Paper

  Github

• Jan 31, 2025
  Placeholder

  Placeholder