The EOSC AIssistant project aims to develop an agentic, Generative AI-powered open research assistant integrated with the European Open Science Cloud (EOSC). It addresses the need to improve the efficiency, reproducibility and trustworthiness of scientific workflows by supporting researchers throughout the research lifecycle while maintaining human oversight over ethics, safety and scientific integrity.
The project will combine agentic AI with FAIR Digital Objects, Research Object packaging and Cognitive Knowledge Graphs to support activities including data discovery, FAIRification, provenance tracking and hypothesis generation. Its modular architecture will integrate general-purpose and domain-specific agents into reproducible workflows orchestrated through EOSC Nodes, providing transparent and provenance-rich support aligned with EOSC requirements.
The approach will be demonstrated in three domains: biodiversity, energy system simulation and materials science. The pilots will address multimodal data harmonisation, synthetic data generation and scientific discovery, while continuous evaluation and human-in-the-loop co-design will be used to assess and refine the developed AI agents. Explainability, auditability and fairness mechanisms will be incorporated into the solutions.
The project will contribute to the development of sustainable and federated AI services within EOSC, supporting more efficient, reproducible and trustworthy research workflows and enabling wider use of AI in European science.
Coordinator: Technische Informationsbibliothek (Germany).
Funding: Horizon Europe
