Welcome!

Hello, I’m Sarvesh Gharat, a Ph.D. student at the Centre for Machine Intelligence and Data Science (CMInDS), IIT Bombay.
My research lies at the intersection of online learning theory and generative AI, with a focus on how we can make intelligent systems learn and decide efficiently under uncertainty.

I work primarily on multi-armed bandits, reinforcement learning, and inference-time algorithms for large language models (LLMs). My recent research explores cost-aware decision-making and model selection for LLMs, aiming to identify and improve the best models for a given task under practical constraints such as compute and feedback cost.

More broadly, I’m interested in LLM alignment and multi-agent systems — understanding how AI agents interact, cooperate, and compete, and how we can design learning algorithms that make these systems more reliable and aligned with human goals. This line of work bridges learning theory, sequential decision-making, and real-world AI deployment.

I have had the privilege of working with outstanding mentors and collaborators, including internships at Google DeepMind and Adobe Research, where I worked on AI for social impact and inference-time reasoning.

Outside research, I enjoy probability theory for its elegance in modeling uncertainty, and I’m an active sports enthusiast — you’ll often find me on the football or cricket field.

I’m always happy to connect with researchers and practitioners interested in online learning, LLMs, and AI systems.
Feel free to reach out!