Pouria Behnoudfar

I am a Van Vleck Visiting Assistant Professor in the Department of Mathematics at the University of Wisconsin–Madison. My research lies at the intersection of machine learning, physical modeling, and data assimilation for dynamical and stochastic systems. I develop and apply these methods to improve the prediction of real-world phenomena, particularly in atmospheric science. At UW–Madison, I am working under the mentorship of Professor Nan Chen.

Before joining UW–Madison in August 2024, I was a postdoctoral fellow at the Commonwealth Scientific and Industrial Research Organisation (CSIRO), Australia. I focused on computational physics and physics-informed machine learning. I completed my PhD under the supervision of Professor Victor Calo, where my work centered on developing stable finite element methods and high-order time-marching approaches for various applications, including phase-field modeling.

My CV can be found here.

My favorite aphorism:

All models are wrong, but some are useful. George Box.

News

August 2026 — IFDS Research Expo: Delivered an invited talk on “Explainable AI for ”
July 2026 — University of California, Davis: Delivered an invited seminar, hosted by Professor Maike Sonnewald, on “Bridging Idealized and Operational Models: An Explainable AI Framework for Earth System Emulators.”
July 2026 — SIAM Annual Meeting: Presented “Ω: Operator-based Mixture Ensemble for Generative Assimilation.”
July 2026 — SIAM Conference on Mathematics of Planet Earth (MPE26): Presented “Bridging Model: Explainable AI for Earth System Emulators.”
July 2026 — Journal of the Mechanics and Physics of Solids: Our paper “SPLIT-PINN: Separable Probability Learning Technique via Physics-Informed Neural Networks for High-Dimensional Probabilistic Modeling” was accepted for publication. [Link]
July 2026 — Journal of Computational Physics: Our paper “RL-DAUNCE: Reinforcement Learning-Driven Data Assimilation with Uncertainty-Aware Constrained Ensembles” was published in the Journal of Computational Physics. [Link]
June 2026: Together with Nan Chen, we submitted “Ω: Operator-based Mixture Ensemble for Generative Assimilation.” [Link]
April 2026 — Courant Institute of Mathematical Sciences, New York University: Presented “Bridging Idealized and Operational Models: An Explainable AI Framework for Earth System Emulators” at the Postdoctoral Seminar.
April 2026 — New Jersey Institute of Technology: Gave an invited seminar on explainable AI approaches for combining physical models of different complexity.
February 2026 — Purdue University: Presented at the STORM Seminar at Purdue University, invited by Professor Di Qi.
March 2026 — SIAM Conference on Uncertainty Quantification (UQ26), Minneapolis, Minnesota: Presented “Uncertainty-Aware Constrained Reinforcement Learning for Data Assimilation.”
2026 — SIAM Annual Meeting: Together with Nan Chen and Charlotte Moser, we will organize the minisymposium “Dynamical Systems and Reduced-Order Modeling for Complex Multiscale Applications.”
2026 — SIAM Conference on Mathematics of Planet Earth (MPE26): Together with Nan Chen, we will organize the minisymposium “Advances in Data Assimilation and Machine Learning.”
2026 — npj Climate and Atmospheric Science: Our paper “Bridging Idealized and Operational Models: An Explainable AI Framework for Earth System Emulators” was published in npj Climate and Atmospheric Science. [Link]
2026 — Inverse Methods for Complex Systems under Uncertainty Workshop: Our position paper was accepted for presentation at the workshop, sponsored by the U.S. Department of Energy, Office of Science, Advanced Scientific Computing Research.
August 21–22, 2025 — University of Wisconsin–Madison: Together with Nan Chen, Jared Whitehead, and Blake Barker, we organized “Data Meets Dynamics: Workshop on Data Assimilation for Complex Systems and Applications.”
2025: Submitted “RL-DAUNCE: Reinforcement Learning-Driven Data Assimilation with Uncertainty-Aware Constrained Ensembles.” Link

Get in touch at behnoudfar [at] wisc [dot] edu

Google Scholar