Trustworthy, FAIR-by-design assistant to help scientists discover data, generate hypotheses, orchestrate workflows, and produce reproducible research outputs inside the EOSC ecosystem.

EOSC AIssistant (Agentic GenAI Open Research Assistant) is a Horizon Europe project funded under HORIZON-INFRA-2025-01-EOSC-05, and coordinated by TIB – Leibniz Information Centre for Science and Technology.

The project will develop an agentic, AI-powered open research assistant integrated into the European Open Science Cloud. It will support researchers across the research lifecycle, including data and literature discovery, hypothesis generation, experimental design, workflow orchestration and the production of reproducible research outputs.

Rather than providing a single scientific chatbot, EOSC AIssistant will combine general-purpose and domain-specific research agents with FAIR data, knowledge graphs, research software and EOSC computational services. Human researchers will retain responsibility for scientific judgement, ethics and critical decisions.

The project will validate the assistant through use cases in:

  • Biodiversity, including multimodal species identification, ecological data integration and conservation monitoring;
  • Energy-system simulation, including scenario design, data preparation and co-simulation;
  • Materials science, including property prediction and optimisation of atomic-layer deposition processes.

Objectives

  1. Improve FAIRness, data quality, provenance and bias detection across research domains.
  2. Develop trustworthy GenAI models for scientific workflows and reasoning tasks.
  3. Agentic AI Assistant for seamless and trustworthy human-AI research collaboration.
  4. Validate the approach through targeted use cases (biodiversity, energy systems and materials science)

RO-Crate and FAIR Digital Objects

RO-Crate provides a common, machine-actionable way to connect research data with its wider context, including people, software, workflows, models, licences and provenance.

Across EOSC AIssistant, RO-Crate and FAIR Digital Objects will support the packaging and exchange of datasets, models, agents, benchmarks, computational workflows and research results. This will help preserve traceability across:

data → model → agent → workflow → research result

RO-Crate profiles will support FAIR data foundations, reproducible agent deployment, workflow execution provenance and governance. The materials-science use case will also use RO-Crate/FDO packaging for curated corpora and reproducible experimental outputs. Detailed provenance is expressed in PROV-O.

eScience Lab involvement

The eScience Lab contributes expertise in RO-Crate, WorkflowHub, scientific workflows, provenance and FAIR Digital Objects. The project has overlaps with AI4Social, CDIF4EOSC, ClimateAdapt4EOSC and earlier FDO work in EuroScienceGateway and BioDT.

UNIMAN contributions

  • WP1: FAIR Data Foundations for Research Workflows (lead: OU)
    • Task T1.3: Metadata schemas and data-sharing infrastructure for FAIR Digital Objects (lead: PSNC)
  • WP2: Adaptive GenAI Models for General-purpose Research Tasks (lead: NKUA)
    • Task T2.4: EOSC-ready deployment of adaptive models (lead: UNIMAN)
    • Deliverable D2.4: EOSC-Integrated Agentic Services (lead: UNIMAN)
  • WP4: Domain-Specific Models and Agentic Assistant for Specialised Research Workflows (lead: PSNC)
    • Task T4.4: EOSC-ready deployment of domain-specific models (lead: PSNC)
  • WP6: Trustworthiness and Bias (lead: LUH)
    • Task T6.1: Bias Identification and Mitigation in Data and Models (lead: UOC)
    • Task T6.2: Evaluation Framework for Trustworthy AI (lead: LUH)
    • Task T6.3: Explainability and Transparency Mechanisms (lead: TIB)
    • Task T6.4: Governance, Compliance, and EOSC Integration (lead: UNIMAN)

  • WP7: Training and Support for Generative AI in Scientific Research (lead: OFFIS)

  • WP8: EOSC-centered community engagement and support programmes for implementing Generative AI in Research workflows (lead: PUPIN)
    • Task T8.1: Project identity and outreach framework (lead: LUH)
    • Task T8.2: Demonstrators for AI-assisted quality assessment and bias mitigation (lead: PUPIN)
    • Task T8.3: Advancing machine-actionable research data and AI-enabled services (lead: LAION)
    • Task T8.4: Protocols and policies for AI-driven workflows within EOSC (lead: UNIMAN)
    • Task T8.5: Advisory Board and Links with Industrial User Groups/Associations (lead: UNIMAN)
    • Deliverable D8.4: Stakeholder and Advisory Board Engagement Report (lead: UNIMAN)
  • WP9: Project Management and Coordination (lead: TIB)
    • Task T9.1: Scientific Coordination (lead: TIB)
    • Task T9.2: Administrative and Financial Management (lead: TIB)
    • Task T9.3: Gender and Data Governance (lead: LUH)