MIT '27 · Computer Science & Mathematics

I'm an undergraduate researcher at MIT CSAIL working on language model reasoning, adversarial self-play, and in-context learning. My research focuses on teaching LLMs to recover from flawed reasoning and adapt to novel domains without fine-tuning.

In-Context Learning for Esoteric Programming Languages · DL4C @ NeurIPS '25
Exploring few-shot learning on unconventional DSLs absent from training data, with a self-scaffolding approach for iterative context building.
[Paper] [Code]

Radar & Satellite Imagery for Detecting Illegal Selective Logging · TCCML @ NeurIPS '25
First ML approach fusing radar + satellite data for deforestation detection. 95.08% accuracy on 134K+ image dataset.
[Paper]

Flawed-CoT: Teaching LLMs to Recover from Reasoning Errors · Preprint
Mixed-CoT training with GRPO optimization for mathematical reasoning robustness.
[Paper] [Code]