Science of AI

In the middle of one of Sweden’s largest research organizations, the center coordinates applied AI research with applications in many research fields. Collaborating with partners on a national and international scale, we are committed to addressing the most pressing contemporary challenges through our research endeavors.  

The goal is to contribute to the international development of knowledge and the research frontier in AI. We do this through scientific publications, postgraduate education and teaching offers. These activities are an important part of ensuring our competence at the international top level. 

Our six research focus areas are: 

• Deep Learning is the area studying the fundamentals of modern-day machine learning, a technique that has enabled recent break-troughs such as generative AI including diffusion models and conversational systems. In our research, we focus on aspects such as privacy, efficiency, and sustainability. In our research activities, we develop state-of-the-art deep learning models for applications within environmental research, the energy sector, process industry, and the public sector, among others. 
• Within Natural Language Processing, we further look into aspects of state-of-the-art machine learning techniques and how they apply to the domain of language. This includes computational efficiency for large language models (LLMs), and how that type of models can be used for pharmacovigilance, the study of adverse effects of medications.  
• Within Computer Vision, we develop solutions where data is structured as images. This includes (but is not limited to) camera inputs from handheld devices or autonomous agents, satellite imagery, infrared cameras, and other spatial sensors. Applications include the modeling and forecasting of stream flows, estimation of wetlands using remote sensing, and estimation of the value of secondhand items in circular business settings. 
• Within Advanced Data Analytics, we study different sources of data through the lens of causal reasoning, Bayesian statistics, and information theory. We approach real-world problems related to anomaly detection, diagnostics, and analysis within application areas such as the telecom industry and solar cell production. 
• Within AI Platforms and Edge, we focus on the infrastructure of today’s AI systems. Achieving industrial grade AI requires an understanding of the infrastructure for AI from cloud via edge-cloud and embedded AI. Our work focuses on the energy and resource efficiency of machine learning, efficient data pipelines, and tools and services for the preprocessing of raw datasets, and the deployment of AI driven solutions. 
• Within AI for science, we utilize AI tools to empower and boost scientific discovery. This includes machine learning for physical simulations, AI for materials, and AI for chemistry, modeling chemical properties of compounds and predicting reaction pathways. 

Even though the focus areas highlight approaches and their applicability to specific tasks and applications, the reality of our research is in most cases driven by the application and the need in each use-case. Solutions are often hybrid solutions that take expertise and techniques from several focus areas at the same time. In our research, both the impact from the application and the scientific impact is equally prioritized, and this can be seen in our showcased projects and scientific publications. 

Some cases we have worked on

Read more by clicking on +