Careers Denmark
Browse opportunities

Research & Development

Postdoc in Machine Learning for Energy Materials - DTU Compute

Danmarks Tekniske Universitet (DTU) Kgs. Lyngby, Denmark Posted 27 Oct 2026 Deadline 22 Nov · 5 days left Employer's career siteTemporaryFull timeSenior specialistEnglish

Summary

Develop probabilistic machine learning and active learning methods to guide experiments on chalcogenide thin films for solar cells. The role includes interdisciplinary collaboration, publishing research, and contributing to teaching and student supervision.

The fields are read by AI from the ad. If the employer filled in a field themselves, we use theirs.

Opportunity summary

Do you want to develop machine learning methods that help uncover how materials are formed and guide the next experiment?

If you are building a career as a researcher in machine learning and are passionate about developing methods for scientific discovery, we offer an excellent opportunity to advance your research. We are seeking a highly motivated postdoctoral researcher to join DTU Compute and develop probabilistic modeling and active learning methods for next-generation solar energy materials.

The position is funded by a Villum Synergy grant within the project “Data-driven design of tunable chalcogenide thin films for next-generation indoor and tandem photovoltaics.” You will join an interdisciplinary collaboration between DTU Compute and the Thin-Film Photovoltaics Group at DTU Electro, working closely with a dedicated experimental postdoc.

Project description

How can we learn from limited experimental data to understand materials synthesis and decide which experiment to perform next? This project addresses this question for antimony-based chalcogenide thin films for indoor and tandem solar cells. Their properties depend on complex interactions between synthesis conditions, which influence crystal structure, defect formation, and ultimately photovoltaic performance.

You will develop probabilistic models that connect synthesis conditions with material properties and device performance, combining experimental data with physical and chemical knowledge. Relevant approaches include Gaussian processes, Bayesian neural networks, and uncertainty-aware representation learning.

Alongside accurate prediction, a central aim is to identify informative descriptors and understand which processing conditions govern bandgap, defects, and recombination losses.

The models will form part of a closed-loop framework in which active learning and Bayesian optimization guide successive synthesis experiments. …

This is an extract. Read the whole advert at Employer's career site