I am a Research Scientist at Meta, in Menlo Park, where my work focuses on designing and implementing machine learning (ML) systems that are both theoretically sound and practically impactful --- bridging low-level hardware, algorithms, and high-level applications. My work includes developing algorithms and frameworks to
- Perform on-device ML inference with minimal memory and compute (EdgeML on microcontrollers),
- Enable efficient federated learning on heterogeneous networks (e.g., k-fed: one-shot federated clustering),
- Support adaptive inference for cloud-based LLM deployments (e.g., B-Distil, ABC),
- Continually solve large-scale optimization problems (e.g., COpter: a system for continual optimization in resource allocation).
Previously, I was a PhD student in the Machine Learning Department at Carnegie Mellon University, advised by Prof. Virginia Smith. In Summer '21, I interned with Kazuhito Koishida at Microsoft Applied Sciences Group, Redmond, where I worked on adaptive ML inference methods. Prior to that, I spent two enriching years as a Research Fellow at Microsoft Research India, advised by Prateek Jain and Harsha Simhadri. I completed my undergraduate studies at the Indian Institute of Technology, Patna, where I worked with Prof. Arijit Mondal on designing a RISC-V microprocessor and accompanying toolchain, applied in FPGA-based hardware simulations and educational tools.