Ph.D. candidate · UC San Diego
I am a final-year Ph.D. student in the Department of Computer Science and Engineering at the University of California San Diego (UCSD), working with Prof. Arya Mazumdar. During my Ph.D., I have interned at AWS Agentic AI, Google Research, and RIKEN AIP.
My research aims to build AI systems that learn efficiently and generalize reliably. I pursue this goal by developing better learning algorithms, establishing their theoretical foundations, and studying how models acquire and transfer knowledge. My current focus is LLM post-training for reasoning models and agents.
I develop reinforcement learning and on-policy distillation methods that help reasoning models and agents learn more effectively from teacher supervision and task feedback. My work aligns token-level learning signals with task rewards, connects distillation to RL advantage estimation, and uses privileged training information to improve agent learning.
I study how data, optimization, and model structure determine learning efficiency and generalization. My work develops generalization bounds through connections between training trajectories and kernel methods, analyzes transformer in-context learning, and establishes guarantees for dataset distillation using compact synthetic datasets.
I investigate what knowledge neural networks acquire and how it is represented and transferred across models and tasks. My work studies relationships between fine-tuned representations, improves automated neuron explanations, and quantifies concepts learned through distillation.
Under review; also accepted at NeurIPS 2026 PretoPost Workshop.
Under review; also accepted at NeurIPS 2026 PretoPost Workshop.
Under review; also accepted at NeurIPS 2026 PretoPost Workshop.
NeurIPS 2023 Workshop on Attributing Model Behavior at Scale.
Jun. – Sep. 2026
Applied Scientist Intern · Santa Clara, USA
Host: Vijay Lingam
LLM Post-training: Reinforcement Learning and Distillation
Jun. – Sep. 2025
Student Researcher · New York, USA
Hosts: Ankit Singh Rawat, Aditya Krishna Menon
On-Policy Distillation for LLM Reasoning
Jun. – Sep. 2024
Research Intern · Tokyo, Japan
Advisor: Prof. Taiji Suzuki
Theory of Neural Networks and Language Models
Sep. 2022 – Mar. 2027 (expected)
Ph.D. in Computer Science
Advisor: Arya Mazumdar
Sep. 2020 – Jun. 2022
M.S. in Computer Science
Aug. 2015 – Jun. 2019
B.E. in Information Engineering