Chinmaya Kausik

• Tackling tangible, real-world questions with a principled mathematical approach. These days, my research focuses on sequential decision making under various settings - offline-to-online transfer, partial observability/latent information and non-standard feedback and reward models. I also have side projects in deep learning. On the other hand, a lot of my undergraduate background was in geometry, topology and dynamics, with work in computer-assisted topology and geometry.
• Increasing accessibility to and in higher mathematics and creating communities where ideas cross pollinate and people pull each other up. I have started the Stats, Physics, Astronomy, Math (SPAM) graduate student social initiative at the University of Michigan. I also co-founded and co-organize Monsoon Math Camp. I have also been involved in building and expanding other mathematical communities, like platforms for the PolyMath REU, DRP programs and the undergraduate math organization at IISc, etc.

• Writing a paper based on my internship at Microsoft in the advertiser optimization team under Ajith Moparthi! I designed and implemented a fast algorithm for updating models used for advertiser bidding.
• Collaborating with Yonathan Efroni (Meta), Aadirupa Saha (Apple), Nadav Merlis (ENSEA) on algorithms for bandit and reinforcement learning algorithms with feedback at varying costs and accuracies, also called multi-fidelity feedback.
• Thinking about principled approaches to data collection and learning for RLHF under real-world considerations.
• Formulating problems in learning under latent information and nonstationarity in bandits.
• Organizing an interdepartmental social initiative, SPAM (Statistics, Physics, Astronomy, Mathematics).
• Fleshing out ideas for more academic communities like Monsoon Math.

primary goals

• Complete an empirical study of RLHF methods on LLMs of varying size and understand the implementation nuances of major RLHF methods.
• Work on a large scale applied recommender systems project using the latent bandit algorithms that I designed (LOCAL-UCB and ProBALL-UCB).
• Applying ideas from RLHF and bandits to mental health studies that my advisor is involved in.

side-quests

• Design a codenames bot using one LLM and train it againts players designed using a different LLM.
• Explore the nuances of implementing various RL algorithms in simulated motion settings.
• Design meaningful experiments to compare LLM agents trained using language feedback with RL agents trained using numerical feedback, using benchmark frameworks like LLF-bench.