Amrith Setlur

I’m a 4-th year PhD Student in the Machine Learning Department at Carnegie Mellon University, where I am fortunate to be advised by Virginia Smith. During my PhD, I have had the privilege of also being a long term visiting researcher at UC Berkeley, advised by Sergey Levine. My PhD is generously supported by the JP Morgan AI PhD Fellowship award.

Research Overview

My research broadly focuses on addressing challenges in modern machine learning (ML) paradigms that remain unsolved after scaling data. As we exhaust internet data, my recent work has focused on enhancing generalization by developing new axes of sustained performance scaling through self‑training, leveraging synthetic data with automated supervision, and optimizing models to use test‑time compute via reinforcement learning. In the past, my work has spanned other challenges of modern ML like robustness to spurious correlations, out‑of‑distribution generalization, and data privacy leakage/memorization.

Collaborations

During my PhD I have a lot of fun collaborating with some amazing researchers across different areas, from robustness, privacy, to NLP, reinforcement learning and more recently LLM reasoning. Amongst the many, I have particularly enjoyed working closely with Aviral Kumar, Aditi Raghunathan, and Vitaly Feldman.