Machine Learning for Dynamic Systems
Neural ODEs, sequence models, and diffusion models for stochastic and high-dimensional dynamics.
AI/ML engineer & Ph.D. researcher
Building learning and control systems for complex dynamics.
I develop machine-learning, stochastic modeling, inference, planning, and control methods for complex dynamical systems in autonomous vehicles, robotics, and energy.
- Machine learning
- Stochastic processes
- Bayesian inference
- Diffusion models
- Optimal control
Based in Michigan
Ph.D. student at Michigan State University
Simulations / Physical AI
Scalable simulation and physics-constrained learning for complex physical and energy systems.
Control and Estimation
Bayesian inference, inverse problems, state estimation, model predictive control, and optimal control.
Autonomous Systems
Motion planning and decision-making for autonomous vehicles and robotic systems under uncertainty.
Current work
Research ideas carried through to working systems.
My work connects model design, large-scale training, quantitative evaluation, and decision-making under uncertainty.
Industry · 2026
Scalable scenario conversion for autonomous-driving simulation
At the General Motors Cruise team, I built a multimodal learning system that reconstructs missing vehicle motion in recorded-driving simulations.
- Approximately 1 TB of source data
- Training reduced from 12 hours to 3
- Evaluation reduced from over 60 minutes to about 9
Research · Optimal inferential control
Bayesian control for learned and high-dimensional dynamics
I reformulate optimal control as probabilistic inference and develop GPU-compatible ensemble methods for neural and spatial models.
View the project
Research · MERL
Physics-constrained neural dynamics for HVAC
A Neural ODE–GRU model improved forecasting accuracy by 14% and ran 5.7× faster than the previous modeling approach.
View the publicationResearch direction
Machine-learning models that can support decisions, not only predictions.
My research uses stochastic processes, Bayesian inference, and diffusion models to formulate inverse, estimation, and control problems for dynamical systems. Current models include neural state-space systems, Neural ODEs, and learned representations of Burgers and Navier–Stokes equations.
I work across machine learning, generative modeling, model predictive control, ensemble Kalman methods, and GPU implementation. The objective is to make these methods usable at the scale of modern learned simulators.
Browse publicationsContact