Chinmay Datar, M.Sc.

Background

• Started my doctoral studies with Hans Fischer Senior Fellow Prof. Wil Schilders and Prof. Felix Dietrich in July 2022
• Doctoral candidate at the Institute of Advanced Study (IAS), TUM in the focus group: 'Scientific Machine Learning'
• MSc. in Computational Engineering at FAU, Erlangen (2018-2022)
  • Engineering Summer Internship at Koshizuka-Shibata Lab, University of Tokyo, Japan
  • Supported by the FAU Graduation Scholarship
• Summer Research Internship at the Hanson Lab, Stanford University, USA (2017)
• Bachelor's degree in Mechanical Engineering from the University of Pune (2017)

Research interests: Scientific Machine Learning

I am developing methods that synergistically combine classical methods in scientific computing with deep learning to solve Partial Differential Equations (PDEs). In particular, I am currently working on the following ideas:  

• Dynamic Neural Networks: Understanding and developing novel neural network architectures for simulating dynamical systems.
• Frozen PINNs: Fast, accurate, and gradient-descent-free training of neural PDE solvers via random feature methods.
  • Domain decomposition for solving multi-scale PDEs
  • Laplace-Beltrami Equation on complicated domains via diffusion maps and Frozen PINNs
  • Weak formulation of PDEs
• SWIM Networks: Developing gradient-descent-free training algorithms (Sample Where It Matters) for neural networks.
  • Fourier Networks
  • Multi-stage networks
• Hamiltonian and Graph Hamiltonian Neural Networks: Rapid training via random feature methods.

Publications

You can find my publications in the Google Scholar. 

Supervision: Theses, Guided Research, and Research Internships

Ongoing Work: 

• Arnad, María (Master's Thesis, 2026): Solving PDEs in the weak form using Weak Frozen PINNs.
• Pansch, Justus (Bachelor's Thesis, 2026): Solving the Laplace-Beltrami Equation on complicated geometries via Frozen PINNs and diffusion maps.
• Grossmann, Aristid (Guided Research Project, 2026): Rapid training of multi-stage and Fourier networks without gradient descent. 

Finished Works:

• Mylarassu, Pranav (Visiting Research Intern, Indian Institute of Technology, Bombay (IIT-B), 2026): Convergence analysis of Frozen PINNs for solving the  Laplace-Beltrami Equation on complex domains.
• Yildiz, Eray (Master's Thesis, 2025): Solving PDEs with SWIM networks using domain decomposition.
  • Paper: ICLR 2026 Workshop on AI and Partial Differential Equations
• Semiz, Ahmet (Master's Thesis, 2025): Predicting fluid dynamics using convolutional random feature models.
• Atamert Rahma (Master's Thesis, 2024): Sampling neural networks to approximate Hamiltonian functions.
  • Paper: NeurIPS 2024 Workshop on Machine Learning and the Physical Sciences
• Aditya Phopale (Master's Thesis, 2024): Solving partial differential equations using neural networks with domain decomposition.

Teaching

Winter Semester 26/27:

• Seminar: Case Studies in Computational Science and Engineering
• Lab course: Scientific Computing Lab

Summer Semester 26:

• Seminar: High-Dimensional Methods for Scientific Computing
• Seminar: Scientists and Ethics

Winter Semester 25/26:

• Lecture: Scientific Computing and Machine Learning
• Lab course: Machine Learning for Crowd Modeling and Simulation

Summer Semester 25:

• Seminar: High-Dimensional Methods for Scientific Computing
• Lab course: Machine Learning for Crowd Modeling and Simulation

Winter semester 24/25:

• Lecture: Scientific Computing and Machine Learning
• Seminar: Scientists and Ethics