My work is on strategy and skill learning for multi-robot systems. I study how a team of robots learns to coordinate through reinforcement learning and game theory, and how to detect and react to another’s intention. This can be a dynamic obstacle or actively adversarial.
I’m happy to discuss research or collaboration. Do reach out!
For anything without a public link, or for a copy of a paper you cannot access, email me.
Preprints & Under Review
Work circulated ahead of peer review. Venue is not claimed until acceptance.
RL²-VLA: Adaptive RL Latent Compositional Steering with Test-Time Scaling for Vision-Language-Action Models
2026 Preprint
An inference-time steering framework that composes a lightweight offline RL policy with a frozen vision-language-action model, activating only when the base policy is likely to fail. On the SIMPLER and PolaRiS benchmarks, out-of-domain success rates improve by up to 17.3 percentage points.
Thesis
A natural language processing approach to improve demand forecasting in long supply chains
Master of Engineering Thesis, Massachusetts Institute of Technology, 2020
A deep learning model (NEMO) that uses natural language processing signals from public text to forecast commodity demand in long supply chains, where information sharing across firms is rarely feasible.