199. Meet Gaia: Building Poland’s Factory for European AI

Abstract:

Europe’s ambition to become the world’s first “AI Continent” rests on a growing network of AI Factories – EuroHPC-linked centres across the continent that combine AI-optimised supercomputing with curated data, domain expertise and support services. Gaia AI Factory is Poland’s newest addition to this network. Coordinated by ACC Cyfronet AGH in Krakow and built by a multidisciplinary consortium spanning five national HPC centres, universities, clinical institutes and technology partners, Gaia will give researchers, companies and public institutions access to AI infrastructure and expertise, while connecting them to the wider European ecosystem.

In this talk I will introduce Gaia: why it is being built, what it will consist of, and what it will actually offer its users. I will cover the planned AI-optimised system and the service layer that sits on top of it. I will pay particular attention to healthcare, one of Gaia’s core domains alongside large language models and space technologies, and to what a GDPR-compliant and AI Act-aligned environment for medical data means in practice.

Finally, I will explain how researchers, clinicians, startups and public institutions will be able to access Gaia’s resources, and where Sano – as one of the consortium partners – fits into this picture.

About the author:

Marek Kasztelnik is a Principal Software Engineer, architect, and team leader at ACK Cyfronet AGH. He specialises in developing solutions that enable scientific applications to run on distributed HPC resources and cloud infrastructures. He serves as the Cyfronet lead for research and implementation projects in the fields of computational medicine (including Gemini) and artificial intelligence (Meetween, DVPS), and acts as the coordinator of Gaia AI Factory. He is the creator of the Model Execution Environment—a data pipeline system for patient simulations integrated with the Helios, Ares, and Athena supercomputers—and served as the architect and lead developer for the initial versions of the EOSC Marketplace and the PLGrid grant system. He has authored over 50 publications and presentations covering e-science, personalised medicine, HPC, and AI.