At NVIDIA Research, I work on Nemotron large language model and vision-language model research, focusing on agentic reinforcement learning. I investigate how model architectures and reasoning-oriented learning objectives interact.
At Princeton University, I contribute to Princeton AI Lab research on LLM reasoning and emerging model capabilities. I examine reinforcement learning, model architectures, and representation learning through machine learning experiments.
At ByteDance Seed, I contributed to research on LLM pretraining and scaling, exploring how model scale and training approaches relate to downstream capabilities. I also worked on language-model pretraining and scaling research during an earlier ByteDance Seed internship.
My research has also included a visiting appointment with the UCLA AGI Lab and graduate machine learning research at Tsinghua University. My selected publications include work on group representational position encoding and KL-regularized policy gradient algorithms for LLM reasoning.

