I'm an MS candidate in Natural Language Processing at UC Santa Cruz, working with Dr. Leilani Gilpin and Dr. Chenguang Wang.

I received my Undergrad from the University of Arizona. I also served as a Research Scholar at the University of Edinburgh, where I worked on Neurosymbolic AI.

My current research centers on reward design and verification in LLM alignment specifically understanding when process supervision beats outcome supervision in reinforcement learning from verifiable rewards (RLVR), and how we can build more reliable training signals for reasoning models. I'm also exploring memory architectures for LLM agents, inspired by sleep-cycle consolidation, to help agents retain and organize knowledge over long horizons.

Previously, I worked on constraint-guided traffic generation for autonomous vehicles building neuro-symbolic systems that satisfy physical validity constraints while generating realistic driving scenarios (presented at NeurIPS 2025).

I also contribute to RLLM, an open-source project advancing reinforcement learning for language models.

If you're wondering whether an AI system's reasoning can be trusted whether it's a self-driving car explaining a split-second decision or a language model showing its work on a math proof that's exactly the kind of problem I'm trying to solve :)