Dortmund, 21st August 2026
As part of a new project, ISAS will develop an AI agent that will, in future, independently support researchers with complex biomedical analyses. This will involve setting up the appropriate computing infrastructure.
Biomedical research nowadays produces huge, complex volumes of data, for example from imaging or omics analytics. Analysing such data sets manually often takes months, whereas an AI agent can complete the same analysis in a fraction of the time. Unlike a single AI model, an AI agent autonomously handles complex multimodal workflows: it plans multiple analysis steps, reconciles interim results and links different analyses together, enabling researchers to identify scientific correlations relating to disease progression and derive new hypotheses.
A meta-agent as an interpreter between humans and systems
To ensure that as many researchers as possible can benefit from these AI agents, ISAS is developing a meta-agent as a dialogue-based interface to the AI agents. In future, researchers will be able to use natural language to describe to this meta-agent the scientific question they wish to address using their data, without having to programme AI workflows themselves. This is intended to facilitate access to and the efficient use of the AI agent for researchers from a wide range of disciplines and to promote the interdisciplinary use of AI agents in biomedicine.
Researchers at ISAS will set up the appropriate computing infrastructure for this purpose: AGGCInf (AI AGent-native GPU Computing Infrastructure). Depending on the sensitivity of the data, researchers will then be able to work with the AI agent with varying degrees of openness and data security. The Federal Ministry of Research, Technology and Space is funding the three-year project with approximately 5 million euros. The project is closely linked to the Leibniz initiative ‘Leibniz Health and Prevention’, which brings together the Leibniz Association’s expertise in health research, prevention and early detection.








