Agentic GenAI Open Research Assistant

The EOSC AIssistant project will deliver an agentic, Generative AI (GenAI4EU) powered open research assistant seamlessly integrated into the European Open Science Cloud (EOSC). It will combine agentic GenAI, FAIR Digital Objects, Research Object packaging, and Cognitive Knowledge Graphs to transform scientific workflows across domains by automating data discovery, FAIRification, provenance tracking, and hypothesis generation, while ensuring strong human oversight for ethics, safety, and scientific integrity. Its modular architecture integrates general-purpose and domain-specific agents, orchestrated into reproducible workflows via EOSC Nodes, providing transparent, provenance-rich, and EOSC-compliant support for all stages of the research lifecycle. Three high-impact domains , biodiversity, energy system simulation, and materials science, will demonstrate the assistant’s capacity to harmonise multimodal data, generate synthetic datasets, accelerate discovery, and ensure trustworthy, reproducible outputs. Fully aligned with GDPR, the forthcoming EU AI Act, and Open Science principles, the EOSC AIssistant embeds explainability, auditability, and fairness checks by design. Continuous evaluation, benchmarked against domain and cross-domain tasks, will refine agentic behaviours through human-in-the-loop co-design, ensuring trustworthiness, adaptability, and federation readiness. With 12 leading partners from 7 countries, the consortium unites excellence in AI, research infrastructures, and domain science to advance European sovereignty in digital infrastructures. By delivering sustainable, federated AI services embedded in EOSC, the project will increase trust, efficiency, and reproducibility in European science, and position Europe as a global leader in responsible, FAIR, and AI-driven research.