High-level specialised application support service in HPC-powered Artificial Intelligence (AI)

The MINERVA project focuses on advancing HPC-AI knowledge across European machine learning (ML) and AI communities, acting as a central hub for cutting-edge European competences in large-scale ML/AI research and development, including training, customization, management and related knowledge (e.g. datasets, procedures, compliance with regulation). At its core, MINERVA is focused on bridging the existing gaps in HPC utilization by ML/AI communities by facilitating the integration of cutting-edge HPC infrastructure in their workload thanks to a rich service portfolio that includes three different support levels, from support for porting AI applications and workflow on HPC infrastructures to support for the pre-training and customization of large-scale open foundation models. Indeed, special emphasis is placed on the support around research and development of open- source foundation models, as these can bolster competitiveness and foster innovation within the European digital ecosystem, ensuring the reuse of high quality models and datasets. MINERVA's comprehensive framework includes targeted training programs, availability of dataset, procedures and software stack in the participating HPC centres, best practice guides, interaction with the AI communities to collect and address their needs, as well as foster the development of a Community Hub that will bring together AI researchers, developers, users and HPC specialist with the aim of sharing knowledge and bring Europe at the forefront of the AI revolution. The project's outcomes are expected to significantly contribute to the strategic objectives of EuroHPC and the European Commission, positioning Europe as a leader in the digital age through the innovative application of HPC and AI technologies in scientific research, industry applications, and beyond.