Prakash Thakolkaran

LEE N205

Leonhardstrasse 21

8092 Zürich

I work at the intersection of machine learning and materials science, focusing on the design, modeling, and interpretation of complex materials across scales. My research develops data-driven methods that respect physical laws and remain interpretable, even when data is scarce. I have worked on inverse design of mechanical metamaterials, constitutive modeling from indirect experimental measurements, and generative models for molecules and polymers under physical and chemical constraints.

My current interests center on large-scale generative and representational learning for structured material systems, including graph-based architectures. I am particularly interested in building models that can transfer knowledge across material classes and length scales, and in understanding what such models learn about structure–property relationships when trained on large, heterogeneous datasets.

news

Oct 01, 2025	Started as a Postdoctoral Scientist at the Mechanics & Materials Laboratory at ETH Zürich.
Sep 08, 2025	Successfully defended my PhD Thesis at the Mechanics, Materials and Computing Group at the Delft Univeristy of Technology.
Oct 22, 2022	Selected as the finalist for the Robert J. Melosh Medal at Duke University.

selected publications

Deep learning reveals key predictors of thermal conductivity in covalent organic frameworks

Prakash Thakolkaran, Yiwen Zheng, Yaqi Guo, Aniruddh Vashisth, and Siddhant Kumar

Digital Discovery, 2025

Bib

@article{thakolkaran2025deep,
  title = {Deep learning reveals key predictors of thermal conductivity in covalent organic frameworks},
  author = {Thakolkaran, Prakash and Zheng, Yiwen and Guo, Yaqi and Vashisth, Aniruddh and Kumar, Siddhant},
  journal = {Digital Discovery},
  volume = {4},
  number = {11},
  pages = {3351--3362},
  year = {2025},
  publisher = {Royal Society of Chemistry}
}

Experiment-informed finite-strain inverse design of spinodal metamaterials

Prakash Thakolkaran, Michael Espinal, Somayajulu Dhulipala, Siddhant Kumar, and Carlos M Portela

Extreme Mechanics Letters, 2025

Bib

@article{thakolkaran2025experiment,
  title = {Experiment-informed finite-strain inverse design of spinodal metamaterials},
  author = {Thakolkaran, Prakash and Espinal, Michael and Dhulipala, Somayajulu and Kumar, Siddhant and Portela, Carlos M},
  journal = {Extreme Mechanics Letters},
  volume = {74},
  pages = {102274},
  year = {2025},
  publisher = {Elsevier}
}

NN-EUCLID: Deep-learning hyperelasticity without stress data

Prakash Thakolkaran, Akshay Joshi, Yiwen Zheng, Moritz Flaschel, Laura De Lorenzis, and Siddhant Kumar

Journal of the Mechanics and Physics of Solids, 2022

Bib

@article{thakolkaran2022nn,
  title = {NN-EUCLID: Deep-learning hyperelasticity without stress data},
  author = {Thakolkaran, Prakash and Joshi, Akshay and Zheng, Yiwen and Flaschel, Moritz and De Lorenzis, Laura and Kumar, Siddhant},
  journal = {Journal of the Mechanics and Physics of Solids},
  volume = {169},
  pages = {105076},
  year = {2022},
  publisher = {Elsevier}
}