PhD Student in Electronic-Structure Machine Learning Materials

The Paul Scherrer Institute PSI is the largest research institute for natural and engineering sciences within Switzerland. We perform cutting‑edge research in the fields of future technologies, energy and climate, health innovation and fundamentals of nature. By performing fundamental and applied research, we work on sustainable solutions for major challenges facing society, science and economy. PSI is committed to the training of future generations.

Therefore, about one quarter of our staff are post‑docs, post‑graduates or apprentices. Altogether, PSI employs 2300 people.

This PhD project is part of the new Swiss project “Learning the electrons:
Design, training and application of a general model of the electronic structure of matter”, which aims to develop next‑generation machine‑learning models for electronic‑structure theory. Building on recent advances in machine‑learned interatomic potentials and electronic‑structure simulations, the project seeks to create transferable and scalable models capable of predicting not only energies and forces, but also advanced electronic properties of materials with high accuracy and efficiency.

The project combines developments in machine learning, quantum‑mechanical simulations, and scientific software infrastructure, and is jointly led by Dr Giovanni Pizzi PSI and Prof Dr Michele Ceriotti EPFL. The goal is to develop and apply machine‑learning approaches that provide an explicit representation of the electronic structure of materials, enabling the prediction of advanced electronic properties beyond standard interatomic potentials. Building on state‑of‑the‑art electronic‑structure methods and modern ML architectures, the project will investigate the design, training, and validation of transferable electronic‑ML e‑ML models across a broad range of materials systems, including approaches based on transferable foundation models for materials and large‑scale ML architectures applicable across the periodic table.

For the Materials Software and Data Group in the Laboratory for Materials Simulations of the PSI Center for Scientific Computing, Theory and Data we are looking for a

• Contribute to the co‑development of transferable e‑ML models, investigating the interplay between model design, training strategies, computational efficiency, transferability, and predictive accuracy across a broad range of materials systems
• Generate and curate high‑quality electronic‑structure datasets using automated and reproducible AiiDA‑based workflows for model training and benchmarking
• Validate and benchmark the predictive performance of the models for advanced materials properties beyond standard band structures and charge densities, including electron–phonon coupling and operators and observables related to Berry phases and other electronic‑structure quantities
• Explore the development of transferable foundation models for materials applicable across the periodic table
• Contribute to the development of robust, reusable, and efficient open‑source software and workflows, integrating machine‑learning frameworks with established electronic‑structure codes

We are looking for a highly motivated candidate with a background in computational materials science or condensed‑matter physics, and a keen interest in developing and applying advanced simulation methods and implementing them in workflows. You have experience working independently but also enjoy working in an interdisciplinary and collaborative environment and are eager to combine methodological development with real scientific applications. We do not expect candidates to be experts in all techniques at the start of the PhD;

training and learning will be an integral part of the project.

Requirements for candidates include:

• Master’s degree (or close to completion) in physics, materials science, chemistry, engineering, or a closely related field
• Hands‑on experience using density functional theory DFT for research or projects, and/or experience in the development of machine‑learning ML models applied to materials
• Working knowledge of Python for scientific computing…