Accelerating the discovery of clean energy materials requires integrating experimental research with machine learning. Self-Driving Laboratories (SDLs) are emerging research environments where experiments are planned, executed, and analyzed in closed-loop workflows that combine automated experimentation with AI-driven decision-making. At DIFFER, in close collaboration with our external partners, we are developing an SDL dedicated to accelerated discovery of functional energy materials.
In this PhD project, you will develop machine learning models that learn from high-throughput experimental datasets to uncover structure–property relationships and guide the selection of new experiments. The datasets will include measurements obtained from automated synthesis and optical/electrical characterization workflows. You will be embedded in the AMD research group, and work closely with experimental collaborators to ensure that model development aligns with data quality, measurement conditions, and evolving research priorities.
The SDL operates as a closed-loop system in which each experiment informs the next. Your models will first be used to analyze completed experiments and identify trends, and later integrated into active learning and Bayesian optimization frameworks to suggest which experiments should be performed next. Through this integration, your work will directly shape the experimental strategy of the SDL and accelerate the discovery of new materials.
This position offers a unique opportunity to conduct research at the interface of machine learning, materials science, and autonomous experimentation, contributing to the development of next-generation approaches for data-driven clean energy research.
Responsibilities - Develop and implement machine learning models to analyze and predict materials properties and performance trends from high-throughput experimental data.
- Design and evaluate feature engineering and data representation strategies for heterogeneous datasets obtained from material synthesis, characterization, and functional testing.
- Apply uncertainty-aware modeling, active learning, and Bayesian optimization approaches to guide experiment selection and support closed-loop decision-making in the SDL.
- Work closely with collaborators to align model development with measurement workflows, data availability, and evolving experimental priorities.
- Ensure reproducible and well-documented analysis practices and contribute to FAIR-aligned data interpretation.
- Explore advanced model families (e.g., generative models or graph/equivariant neural networks) to accelerate candidate discovery and hypothesis generation.
- Disseminate research findings through publications, conference presentations, and consortium meetings.
- Supervise junior student projects where appropriate.
- Complete and defend a PhD thesis within four years.