Felix Dietrich

Prof. Dr. Felix Dietrich holds the Professorship for Physics-Enhanced Machine Learning at the Technical University of Munich (TUM), within the TUM School of CIT and Department of Computer Science. He is actively engaged in teaching courses including Scientific Computing and Machine Learning for the Winter semester 2025/26 and supervises Master's theses in Machine Learning in Crowd Modeling & Simulation.

His research focuses on the intersection of scientific computing and machine learning, particularly through his 'Harmonic AI' framework that bridges linear operator theory with deep learning. Key research areas include physics-enhanced machine learning, Koopman operator theory, scientific machine learning, and numerical algorithms for complex systems. His work applies to diverse domains such as crowd dynamics, quantum systems, molecular dynamics, and civil engineering applications like structural defect detection.

Prof. Dietrich leads several major research initiatives including the Emmy Noether group 'Harmonic Artificial Intelligence based on Linear Operators', the Scientific Machine Learning Focus Group with the IAS, GNI projects DeepMonitor and FORWARD, and AutoMD-AI for molecular dynamics simulations. His recent publications demonstrate strong activity in connecting mathematical rigor with machine learning applications.

His scientific contributions span developing data-driven algorithms that combine the reliability of traditional scientific computing with the flexibility of AI approaches. This work has resulted in numerous publications in top-tier journals including SIAM Journal on Scientific Computing, Chaos, Nature Communications, and PNAS Nexus.

• Emmy Noether group on Harmonic AI
• Scientific Machine Learning Focus Group with IAS
• GNI project DeepMonitor for structural defect detection
• GNI project FORWARD for pedestrian flow prediction
• AutoMD-AI for molecular dynamics surrogate modeling

Prof. Dietrich actively mentors PhD students and postdoctoral researchers, with several advisees including Chinmay Datar, Erik Lien Bolager, and Vladyslav Fediukov appearing as co-authors on recent publications. His research group (SCML) develops open-source software for physics-enhanced machine learning applications.