Research Theme

Generative Models for Physical Systems

We build physics-grounded generative models to accelerate discovery in fluid mechanics, heat transfer, sediment transport, and Earth-system dynamics — using AI not as a black box, but as a partner to theory and simulation.

Turbulence & Flows Heat Transfer & Cooling Probabilistic Data Assimilation

In our work, AI is tightly coupled to physical reasoning.

We use generative and probabilistic models to:

• learn distributions of fields (velocity, temperature, concentration)
• emulate and accelerate high-fidelity simulators
• quantify uncertainty and extremes in mechanical and Earth systems
• inform design and control for engineered systems

Rather than replacing equations, we embed physics inside machine learning.

• How can generative ML emulate or augment ensembles for turbulent flows and heat transfer?
• Can symbolic ML recover interpretable transport and scaling laws?
• How do we combine data assimilation, generative models, and physical constraints?
• What is the most efficient way to design cooling strategies for large data clusters?