Research projects within AI4Research
Four research projects associated with artificial intelligence and machine learning have been awarded funding within AI4Research 2025. Read more about the projects below.
AI-accelerating atomistic simulations for green materials discovery
Anders Bergman, Department of Physics and Astronomy
To enable a green transition and reduce our global environmental and climate footprint, new functional magnetic materials are needed. A promising approach to discovering these materials is through so-called digital alchemy – using computer simulations to predict new materials and their properties. For these simulations to produce reliable results, realistic models are required. However, one challenge is that the more accurate the simulations, the greater the computational power and time they require. Therefore, it is crucial to improve simulation methods, so they become faster and more efficient without compromising reliability.
In this project, we will develop new, efficient methods for simulating magnetic materials. We combine traditional calculations with AI-based algorithms, such as machine learning, to achieve this goal. The aim is for these improved methods to be used to identify new magnetic materials such as permanent magnets and magnetocaloric materials, that can serve as components in future green technologies.
What do you look forward to the most during your sabbatical?
To be in a dynamic environment where I can engage with new ideas and perspectives on AI and build valuable connections. I believe this will provide me with both inspiration and new tools for my future research.
Enhancing Cybersecurity through Robust Machine Learning
Christian Rohner, Department of Information Technology
The project focuses on enhancing cybersecurity by improving the robustness of machine learning models to better detect and mitigate evolving cyber threats. It addresses key challenges such as the scarcity of attack data and the need for models to adapt to unseen threats and adversarial attacks. The research is structured around three core aspects: (1) Enhancing Training Data through synthetic data generation and simulation, using techniques like Generative Adversarial Networks (GANs) to address imbalances and improve generalization; (2) Developing Robust Model Architectures by optimizing configurations and employing methods such as ensemble models to balance performance and resilience; and (3) Exploring Domain Adaptation and Feature Engineering to generalize models across environments, enabling reliable adaptation to new contexts and efficient knowledge sharing.
What do you look forward to the most during your sabbatical?
Inspiration and time to integrate AI based methodologies into my research and educational activities, while collaborating with peers from diverse disciplines who are also focusing on AI.
A self-driving lab for exploration & optimisation of optoelectronic materials
Jonathan Staaf Scragg, Department of Materials Science and Engineering
My objective is to discover higher-performance or more sustainable materials for solar cells and related energy conversion technologies. To do this, we need to identify candidates, develop synthesis processes for them, and experimentally verify their properties to test their technological potential. The major challenge of developing a new material is that we are operating in a high dimensional experimental space which must be first explored – to find suitable conditions for growing samples – and then optimised, to obtain best performance.
This is a task ideally suited for AI/ML methods, but these methods need data, and experimental data are usually sparse are time- and resource- expensive to obtain. In this project, we are constructing a ‘self-driving’ or autonomous laboratory setup which can produce large numbers of samples – and thus, large amounts of data, on thin film optoelectronic materials. This is enabling us to build active-learning based exploration/optimisation strategies for synthesis development, where the experimental conditions for each new sample are selected based on a Bayesian optimisation strategy. The ongoing development of the autonomous hardware as well as AI/ML methods at multiple levels in the workflow are aspects I am focussing on during the AI4Research sabbatical.
What do you look forward to the most during your sabbatical?
As an experimental materials scientist, using AI/ML methods is a new step for me. There is a lot I need to learn, and collaboration will be critical for success. Spending more time with AI specialists, learning from exposure to different methods and ideas, and meeting other scientists in the same position as myself will be a valuable experience. In particular, the discussion of how to couple domain knowledge into more generic AI/ML approaches is of great interest to me.
Machine learning for analysis of stochastic electrochemical processes
Alina Sekretareva, Department of Chemistry – Ångström
Electrochemical processes are inherently stochastic, yet traditional measurement techniques often smooth out their discrete nature due to averaging over numerous events. Single-entity electrochemistry (SEE) methods, however, enable the study of individual systems, revealing stochastic behavior as a key feature.
Manual analysis of SEE data is time-intensive and prone to human error, limiting the exploration of rare events. Machine learning (ML) approaches offer a solution, enabling efficient and unbiased data processing. Building on a previously developed unsupervised ML algorithm for automating SEE data analysis, this project aims to integrate supervised ML techniques into the data analysis workflow to extract physically meaningful information about electrochemical processes. This integration will enhance the understanding of recorded signals, providing valuable physical insights.
What do you look forward to the most during your sabbatical?
I look forward to exchanging ideas on using AI in research with colleagues from diverse backgrounds. This collaborative environment promises to be both intellectually stimulating and transformative for my work.