Cooperation of Artificial Intelligence Factories and Factories Antennas

Expected Outcome

Upon completion of the Action, the European HPC and AI ecosystems will be strengthened through an effective network of AI Factories supporting the adoption and use of HPC in the development of trustworthy artificial intelligence (AI) by startups and SMEs, but also by the private and public sector in general, taking into account the specific needs of the local and national ecosystems. The coordinated network will facilitate synergies and assets reutilisation, support, training, staff exchange, knowledge transfer between, AIF+As, as well as prevent duplication of efforts.

The Action will ensure the network of AIF+As will be embedded in an enhanced European AI/ HPC ecosystem with strong links to other European HPC, AI, and data initiatives (see above).

Moreover, the Action will result in:

• Contribution to the realisation of the EuroHPC overall and specific objectives.
• A common governance baseline across AIF+As to ensure the full interoperability and the collective compliance or the network of AIF+As.
• Seamless user experience across AIF+As, with users receiving a consistent offer of core services.
• Effective coordination and exchange of best practices and information among the network of AIF+As.
• Establishment of a network of AIF Data Labs in 7-8 strategic domains, including a common framework for data access and data management.
• Easy access across AIF+As to up-to-date, rich, high-quality open web data compliant with EU regulations and values.
• Curated access to services and facilities offered by AIF+As.
• Maximised visibility and outreach of AIF+As, in particular to AI startups, SMEs and industry.
• Improved coordination and increased availability of training activities across AIF+As and within the European HPC ecosystem.
• Contribution to the attraction of HPC/AI talent and development of a distributed pool of experts in Europe.

The JU considers that proposals requesting a contribution from the EU of up to EUR 12.5 million and a duration of 3 years would allow this specific challenge to be addressed appropriately, with the following indicative EU budget distribution per subtopic:

• General coordination and networking: EUR 2.5 million
• Networking of AIF Data Labs: EUR 7.5 million
• Provision of EU open web data: EUR 2.5 million

Nonetheless, this does not preclude submission and selection of proposals requesting another duration or other amounts. Only one proposal, covering all three subtopics in the scope, will be selected.

Scope

A. General coordination and networking

Proposals should aim at coordinating and promoting networking and collaboration of the AIF+As. They are expected to establish a communication platform, facilitate dialogue, enable asset sharing, promote the objectives of the AIF+As, and organize outreach events and workshops on topics of interest to the AIF+As and their communities.

The Action will support and enhance the alignment of AIF+As through targeted activities, building common standards of service to provide a harmonised experience to users. The activities should leverage on synergies and complementarity of the AIF+As. It is expected to identify solutions and tools available from the AIF+As network to support and assist AIF+As in addressing requests and needs of their constituencies.

The Action should:

• Assist the development of the AIF+As and coordinate their collaboration, ensuring a seamless user experience across all facilities. Coordinate the joint activities and exchange of best practices across the AIF+As, including the sharing of assets and knowledge to prevent duplication of efforts and speed up developments, and support projects spanning across two or more AIF+As or federating/distributed learning and inference when applicable.
• Attract new European user communities and support the engagement of startups, industry, and SMEs in AIF+As activities, while maximizing visibility and outreach to these groups.
• Promote joint training offerings and the exchange of training materials and courses. Support talent detection, attraction, and development, and enhance mobility of HPC/AI specialists between communities, academia, public, and private sectors.
• Implement and coordinate technology transfer activities at the European level and for the Digital Single Market, and promote the adoption of developed methods and technologies by AIF+As users and the wider European HPC/AI communities.
• Develop a comprehensive AIF service directory, detailing all services offered by AIF+As, including both HPC and cloud-based solutions, as well as associated support services, and advise and support AIF+As with the development of sustainability.
• Support an Annual European AI Factories event connecting all the AI community in Europe, in collaboration with the EuroHPC JU, and promote the networking of AIF+As users, especially for different users’ profiles and sectors, to foster innovation.
• Identify and collect meaningful qualitative and quantitative common KPIs for AIF+As to measure the impact of this initiative on the European HPC and AI ecosystems.

B. Networking of AIF Data Labs

Data Labs contribute to the objectives of the European Data Union Strategy by scaling up access to data for AI. They create the link between data holders, Common European Data Spaces[1], domain-specific data ecosystems, and the AIF+As and the AI innovation ecosystem. Their role is to facilitate the availability and use of high-quality data under appropriate technical, governance, and regulatory conditions in close collaboration with relevant EU initiatives. AIF Data Labs are operational components within the AIF+As that will provide AI developers with access to technical infrastructure, data management tools, and large datasets required for the development, testing, and validation of AI models.

Each Data Lab will offer a consistent set of services, including data discovery, standardisation, cleaning, enrichment, and synthetic data generation, as well as guidance on data governance and compliance with EU legislation. Data Labs will also play a key role in supporting legal and regulatory compliance by providing services such as pseudonymisation and anonymisation of datasets, the use of secure processing environments, and legal assistance on the use of data.

Data Labs will be implemented across a set of priority sectors aligned with those identified by the Apply AI Strategy as having high potential for the development and deployment of trustworthy and impactful AI solutions. These include healthcare and life sciences, manufacturing and robotics, public administration, cybersecurity and internal security, culture and languages, scientific research, and climate and environmental modelling.

The Action should:

• Support the networking and federation of Data Labs across AIF+As into a common European framework, with a strong emphasis on the use of the Simpl open-source middleware as the core interoperability platform between the different data facilities involved in each Data Lab. This framework should ensure interoperability, secure data exchange, and federated access across the AIF+As, while connecting Data Labs to the corresponding Common European Data Spaces[2], and AI flagship initiatives in line with European priorities.

• Enable efficient data use across sectors and borders, ensure regulatory and technical alignment, and promote the reuse of shared tools and resources.
• Integrate Data Labs activities with AIF+As, ensuring that AI developers can seamlessly use datasets and tools provided by the Data Labs in model development and testing.
• Enable the exchange and reuse of data management and processing tools, including for data discovery, cleaning, enrichment, and synthetic data generation.
• Develop legal and regulatory compliance-enabling services within Data Labs, including mechanisms for pseudonymisation and anonymisation, the provision of secure and compliant data processing environments, and guidance on the lawful use and sharing of data.

C. Provision of EU open web data

Proposals should develop, deploy and operate across AIF+As a European federated web data service to ensure sovereignty in the open web data (OWD) independently of external sources.

The Action should:

• Develop services and best practices around open web data for training and fine-tuning of AI models, AI applications, and AI-based search.
• Deploy and operate a web data service, encompassing general/focused crawling to generate multi-modal raw data (text, image, audio, video) covering all EU languages, metadata creation, indexing, searching, and use case partitioning into domain-specific data pools.
• Collaborate with existing EU initiatives providing web data services.
• Integrate the EU open web service into the AIF Data Labs ecosystem.

[1] Including the Simpl platform supporting data access and interoperability among European data spaces.

Including EOSC to facilitate research data access across the EU.

[2] Including the Simpl platform supporting data access and interoperability among European data spaces.

Including EOSC to facilitate research data access across the EU.