Technical founder and research engineer. I work on reinforcement learning for reasoning systems - specifically, how agents learn when to retrieve, when to reason, and when to hypothesize.
My thesis at Cornell introduced search-augmented RL for legal reasoning, achieving a 10.5pp improvement on LegalBench through tool-assisted training with GRPO.
I founded Aeve, an AI research foundry.
| Year | Work |
|---|---|
| 2025 | Interpretable Search-Augmented Reinforcement Learning for Legal Reasoning — MS Thesis, Cornell University. Advisor: Claire Cardie. · ArXiv (coming soon) · Code |
Research interests: post-training (RLHF/GRPO), search-augmented reasoning, interpretability, production ML systems.
| Aeve | AI research foundry. Research, engineering, and product brought together to build intelligence infrastructure. aevelabs.co |
| LexLens | Legal AI research output. Interpretable legal reasoning with constitutional safety. lexlens.app |
| MeetViro | Autonomous meeting agent with social intelligence. Built by the Aeve team. meetviro.com |