Sujai Hiremath

I am an ORIE PhD student at Cornell Tech, NYC. I am motivated by a curiosity of the world. I work primarily on (1) developing robust and sample efficient causal discovery algorithms, and (2) their application in causal inference and reinforcement learning; I am also currently exploring (3) causality for LLM interpretability. Previously, I explored the intersection of causality, LLMs, and RL as a research intern at Amazon Science in Tübingen, Germany. I am advised by Kyra Gan.

I believe causality will be essential to the efficient and safe development of AI systems that have prediction, control and counterfactual reasoning capabitilies that can generalize to unseen environments. I aim to close the gap between theory and practice to enable machine learning systems to truly understand our causal world.

Google Scholar| Contact | LinkedIn | CV

Drafts

From Causal Structure to Efficient Representations: Deep Reinforcement Learning with Causal Rank Regularization
Sujai Hiremath, Shiva Prasad Kasiviswanathan, Dominik Janzing, Patrick Blöbaum, Elke Kirschbaum, Kyra Gan, Krishna Balasubramanian.

From Guess2Graph: When and How Can Unreliable Experts Safely Boost Causal Discovery in Finite Samples?
Sujai Hiremath, Dominik Janzing, Philipp Faller, Patrick Blöbaum, Elke Kirschbaum, Shiva Prasad Kasiviswanathan, Kyra Gan.

Publications

When Additive Noise Meets Unobserved Mediators: Bivariate Denoising Diffusion for Causal Discovery.
Dominik Meier, Sujai Hiremath, Promit Ghosal, Kyra Gan.

LoSAM: Local Search in Additive Noise Models with Mixed Mechanisms and General Noise for Global Causal Discovery.
Sujai Hiremath, Promit Ghosal, Kyra Gan.

Hybrid Top-Down Global Causal Discovery with Local Search for Linear and Nonlinear Additive Noise Models.
Sujai Hiremath, Jacqueline Maasch, Mengxiao Gao, Promit Ghosal, Kyra Gan.