Community Centres of Excellence

Expected Outcome

Community CoEs must demonstrate scientific and technical excellence while ensuring impact at the wider European HPC community including the European industry and/or academia. Proposals should carry out a well-chosen subset (not necessarily all) of the following actions:

• Identify Targeted Applications & User Communities. Clearly define the software codes to be developed, enhanced or optimised, their ownership/licensing schema, and the impact on European HPC users, ensuring a broad and diverse innovation ecosystem. Describe the European user communities of the targeted applications, the current and predicted use on EuroHPC infrastructure/AI Factories as well as the impact of the planned developments on the European users.
• Deliver Highly scalable, optimized codes for Exascale or post-exascale/ advanced AI. CoEs should address frontier technical HPC/AI challenges, including load balancing, resilience, heterogeneity, and engage with AI-driven developments, ensuring scientific applications are fit-for-purpose in the AI and post-exascale era. CoEs should ensure co-design with HPC hardware and software developers and providers.
• Include activities to improve the energy efficiency of applications, algorithms, methods, libraries, and/or tools.
• Training & Capacity Building. Contribute to the overall training and skills activities, jointly with AI Factories, HPC National Competence Centres to address the skills gap and enable cross-sector AI-HPC adoption.
• Maximize Impact through Collaboration. Proposals should– where relevant and appropriate - establish strong links with: • transversal CoEs, Lighthouse Codes;
• the Support Centre for HPC-powered AI Applications, ensuring mutual benefit between HPC code developers and real-world users;
• national and EU-funded projects, fostering technology transfer and best practices at the European level.

• Connect the Community CoE to national scientific communities inviting them to get involved in the activities developed and use their services.

Objective

Community Centres of Excellence (CoEs) are advancing the transition towards more performant and scalable codes, up to exascale and post-exascale including advanced AI capabilities by developing or scaling up existing parallel codes, resulting into effective applications to solve scientific, industrial or societal challenges and addressing the needs of the user communities, advancing high-performance computing (HPC) & AI applications that operate at the frontier of technology, and providing cutting-edge capabilities for the European HPC user community.

These Community CoEs will play a strategic role in pushing applications code to the next level, where possible up to exascale and post-exascale level including advanced AI capabilities. Proposals must demonstrate the key position of their application codes in their communities, describe their current status of development (e.g., performance, scalability, portability, etc.), and convincingly present their strategy and path for improvement towards their targeted next level. They should also describe how the related gain translates into a positive impact in their community. These CoEs will work closely with other HPC stakeholders e.g. relevant high-performance computing (HPC), scientific, and industrial communities to (i) enhance application performance and exploit advanced computing capabilities, (ii) develop and scale up existing application codes towards the next level up, reaching exascale performance, (iii) deliver tangible benefits for scientific and industrial challenges, (iv) provide user-focused and inclusive support, enabling new and underrepresented communities to leverage HPC/AI infrastructures effectively, (v) deliver training activities on the actions developed within the scope of the CoEs, (vi) ensure multidisciplinary collaboration where relevant in order to integrate expertise in application domains, HPC systems, software, and algorithms.

Scope

This call builds upon and complements HORIZON-EUROHPC-JU-2023-COE-03-01 and HORIZON-EUROHPC-JU-2023-COE-01-01 and should associate where relevant with the following actions: HORIZON-JU-EUROHPC-2026-COE-LH-01-02 and HORIZON-JU-EUROHPC-2026-COE-LH-01-03.

The call is expected to support:

• The development of HPC-ready applications with clearly defined objectives and a clear roadmap for bringing the targeted applications to the next level of maturity and scalability.
• A user-driven approach, connecting CoEs with developers, HPC users, industrial and scientific communities.
• Ensuring coordinated efforts with EuroHPC, Artificial Intelligence (AI) Factories, and national HPC programs.

Community CoEs should clearly define the targeted application(s), user communities, and performance needs, while presenting a detailed development plan with key performance indicators (KPIs), milestones, and deliverables and show European added value.

All Community CoEs should be driven by user needs and specifically target European users beyond the CoE consortium to create wider impact. CoEs should be inherently committed to co-design activities (e. g. in collaboration with any relevant transversal CoE) to ensure that future HPC architectures are well suited for the applications and their users (both from academia and industry), providing a high performance and scalable application base.

CoEs should federate existing resources around Europe, exploiting available competences, and ensuring multidisciplinary (combining application domain and HPC system, software and algorithm expertise) and synergies with national/regional programmes.

CoEs should further enlarge and expand these capabilities all over Europe, in particular by including user communities from EU widening countries and countries associated to Horizon Europe that are members of the EuroHPC Joint Undertaking currently developing and advancing their HPC infrastructure and ecosystem.

Proposals should be able to articulate clearly the scientific grand challenges (e.g. fusion energy, defeating cancer, storage battery, or AI-driven cybersecurity) which will be addressed by the applications and justify the advanced HPC performance needs.

Proposals should also develop synergies with preceding and existing CoEs where relevant. Should the proposed work target an area or domain already covered by former or existing CoEs, proposals must clearly elaborate on how their proposal further expands beyond previous work (e.g., new codes, codes upgraded/enhanced or ported to exascale/post-exascale/AI-optimized domains, new user communities, etc.), and avoid any kind of work duplication/overlap.