Speaker
Description
The emergence of AI Factories and domain-specific data spaces is transforming how scientific communities access, share, and exploit research data. Across domains such as health, environmental sciences, earth observation, materials science, digital humanities, and engineering, enabling AI-driven research requires more than advanced computing infrastructures. It requires a data ecosystem capable of supporting discoverability, interoperability, governance, compliance, and the preparation of data for AI applications.
IT4LIA, Italy's AI Factory initiative, is establishing an ecosystem that combines advanced computing resources, AI services, and training activities to support AI adoption by researchers, public administrations, startups, SMEs, and industry. Within this framework, data services provide the mechanisms needed to discover, document, govern, and prepare datasets for AI-enabled research and innovation.
This presentation discusses the experience gained in developing data services and governance capabilities across diverse scientific domains. The talk focuses on the practical challenges of making research datasets discoverable, accessible, and reusable while respecting legal, ethical, and organizational constraints. We examine key capabilities including dataset onboarding and curation, metadata and discoverability services, data stewardship, AI-readiness assessment, and governance frameworks supporting data quality, interoperability, traceability, and responsible reuse.
The presentation highlights practical lessons learned from the development of the IT4LIA data ecosystem. It discusses how challenges related to dataset discovery, onboarding, metadata quality, governance, interoperability, and data access often represent significant barriers to the adoption of AI. These experiences provide a complementary perspective to discussions on secure data processing environments by focusing on the preparation, documentation, governance, and lifecycle management of data before it can be effectively exploited by AI applications.
Similar challenges are being addressed across the broader ecosystem of the Italian National Centre for HPC, Big Data and Quantum Computing (ICSC), including initiatives such as ECHO-TWIN, which is exploring AI-enabled digital twin ecosystems and federated Edge-Cloud-HPC infrastructures across domains such as health, environment, and smart territories. Collectively, these experiences highlight the growing importance of integrated data and computing ecosystems as a foundation for trustworthy, scalable, and reusable AI in science.