Honored to receive the 2026 Scientific Excellence Award from the DAF–MIT AI Accelerator for GEOPROBE, our physics-informed probabilistic surrogate for high-speed aerodynamic design.
Research foundations
- AI Algorithms Generative models and reinforcement-learning post-training.
- Mathematical Foundations Numerical methods, optimization, and physical structure.
- GPU Systems & Performance GPU kernels and distributed ML infrastructure.
About
I am a Ph.D. candidate in Computational Science and Engineering at MIT, co-advised by Prof. Alan Edelman and Dr. Chris Rackauckas at the Julia Lab, MIT CSAIL.
My research focuses on improving foundation models, including LLMs and generative models, by incorporating mathematical and physical constraints into how they learn and generate. I explore this at different stages of the learning pipeline: probabilistic learning with hard constraints, test-time steering of pretrained generative models, and LLM post-training with continuous, physics-verifiable rewards. Scientific simulation provides a natural setting to develop and test these ideas.
I’m also a bit of a performance nerd. My work ranges from GPU-accelerated numerical solvers and optimization methods to improving the efficiency of large-scale LLM post-training. I particularly enjoy problems where rethinking the algorithm and its implementation can make better use of modern hardware.
Previously, I worked at AWS AI Labs on probabilistic learning with hard constraints, and at NVIDIA on kilometer-scale climate emulation and high-performance RL post-training infrastructure for mixture-of-experts language models.
News
Publications
Selected work across AI for science and engineering, computational mathematics, and high-performance ML systems.
Loading publications…
Experience
Ph.D. Research Intern
Developer Technology (DevTech), NVIDIA Santa Clara, CA
Built high-performance RL post-training methods and GPU kernels achieving bitwise true-policy for MoE LLMs in Megatron-LM and NeMo-RL on NVIDIA Blackwell GPUs. Mentors: Daniel Galvez and Aurelien Chartier.
Research Intern
Developer Technology (DevTech), NVIDIA Earth-2 Santa Clara, CA
Led first-author work on multi-resolution diffusion for global climate emulation at 5 km resolution, unifying coarse generation and super-resolution through a shared probability-flow formulation. Mentors: Akshay Subramaniam and Noah Brenowitz.
Applied Scientist Intern
AWS AI Labs, Amazon Science Santa Clara, CA
Developed a differentiable probabilistic projection layer for learning distributions with exact hard constraints, including custom optimization and implicit differentiation. The work appeared at ICLR 2026. Mentors: Danielle Robinson, Bernie Wang, and Michael W. Mahoney.
Now
Research Assistant
Julia Lab, MIT CSAIL Cambridge, MA
Research on LLM post-training with physics-based rewards, geometry-aware neural surrogates, physics-constrained generative models, and high-performance GPU differential-equation solvers. Co-advised by Alan Edelman and Chris Rackauckas.
Research Intern and Consultant
Julia Computing Inc. Remote
Developed pseudo-transient methods for differential equations with CUDA support and automatic differentiation. Mentor: Chris Rackauckas.
Research Intern
Developed slot-identification algorithms and BERT models for conversational AI.