Shishir Sharma

I am a Ph.D. student in Computer Science at McGill University and Mila, advised by Doina Precup and Theodore J. Perkins. My research focuses on reinforcement learning (RL), especially multi-agent RL and model-based RL.

More broadly, I am interested in using multi-agent systems as a conceptual lens for RL: studying how ideas developed for interacting decision-makers can inform learning, adaptation, and generalization even outside explicitly multi-agent environments.

A current focus of my work is fluid-agent reinforcement learning, where the number and identity of agents may change over time. I study both the game-theoretic structure of such settings and practical MARL algorithms in JAX-based environments.

Research themes

• Fluid multi-agent reinforcement learning: learning and coordination when agent populations are not fixed.
• Game-theoretic structure in MARL: equilibrium reasoning, coordination, and decision-making across changing teams.
• Model-based reinforcement learning: planning under uncertainty and compact probabilistic approximations.
• Environments and evaluation: JAX-based MARL environments for spawning, foraging, predator-prey, and migration-style dynamics.