I am a Postdoctoral Researcher at Microsoft Research India, working broadly on designing efficient and reliable retrieval systems. Prior to Microsoft, I completed my PhD in the Department of Computer Science at Duke University, advised by Pankaj K. Agarwal. I also spent the summer of 2024 as a Visiting Scientist at Cornell University, collaborating closely with Sainyam Galhotra.

During my PhD, I designed algorithms with provable robustness guarantees under noise, uncertainty, and incomplete information. In particular, I developed techniques for effectively using weak machine learning models (noisy but inexpensive) in tasks such as clustering, achieving performance comparable to stronger models (more accurate but expensive).

My current work applies these principles to develop agentic retrieval pipelines based on lightweight but less capable models, such as small language models (SLMs) and decision models (e.g., Jev and Microsoft-Decision-1). Specifically, I am developing methods that combine reinforcement learning with techniques from robust algorithm design to make these systems dependable at scale. 

My work has appeared in leading venues across machine learning, theoretical computer science, and databases and data mining, including NeurIPS, ICLR, COLT, SODA, ICALP, KDD, and PODS.