Skip to main content
PDC Center for High Performance Computing

AI Sweden Initiative

Fredrik Viksten & Niclas Fock, AI Sweden

AI Sweden Infrastructure Network Group

AI Sweden is the Swedish national centre for applied artificial intelligence (AI). It is supported by the Swedish government, as well as the public and private sectors across the country. The mission of the initiative is to accelerate the use of AI for the benefit of Sweden, both from the perspective of everyone living in Sweden, and in terms of Sweden in relation to the rest of the world. Since its inception, AI Sweden has gathered more than a hundred partners from all parts of the Swedish AI ecosystem: authorities, regions, municipalities, academia and institutes, start-ups and small to medium-sized enterprises (SMEs), as well as large national and multinational corporations.

To achieve its mission, AI Sweden runs projects of national interest in areas such as information- driven healthcare, AI solutions for the Swedish language, data-driven journalism, and AI to help tackle climate change. As competence is key, AI Sweden also facilitates targeted training provided by, and designed for, its partners, and also enables training for the general public, in particular via an organization known as the Data Factory. This organization was initially established physically in Gothenburg and, over the last year, it has been expanding and spreading across more sites. The Data Factory enables AI Sweden partners to make their data available to others and to access existing data produced by others, as well as to make use of computing power and storage capacity early in their key AI projects.

AI Sweden is not a research institute, but rather facilitates the entire AI ecosystem, and therefore wants to contribute to a culture of sharing, cooperation, and action between partners in Sweden. All the project’s initiatives are undertaken in collaboration with the AI Sweden partners. AI Sweden is funded by Vinnova (the Swedish government innovation agency) as well as by its partners. The organization has around forty employees working at its offices in Stockholm, Gothenburg, Lund, Örebro, Luleå, Sundsvall, Skellefteå, Linköping and Karlskrona.

Why an Infrastructure Network Group?

The ecosystem of the storage and compute infrastructure is growing across Sweden. For those working within AI-related businesses, this is obviously very positive and an important foundation for developing data factory solutions and accelerating the use of AI nationwide. As a response to the growing need to share learning and knowledge about this topic and to boost collaborative efforts, in spring 2021 AI Sweden decided to create an expert group for infrastructure know-how among Swedish AI computer centres.

The purpose of the infrastructure network group is to share knowledge and experience relating to how to build and operate larger, multi-user AI centres for research and development or for providing services. The network group works to identify common challenges, needs and interests. It will accumulate knowledge in the domain in order to accelerate the use of secure, efficient operations. The idea is to increase the possibilities for collaboration between partners working on similar projects or already owning an infrastructure, and to connect key resources across Sweden.

An initial meeting was held late in spring 2021, and the network group now holds regular network meetings about once a month. Quite early on, the group itself decided to limit the scope and member focus to publicly-funded AI data centres in Sweden. AI Sweden has several private corporate partners, and a parallel group for these partners would be considered if su cient interest in a similar network forum arises. External to the expert group(s), there is an infrastructure ecosystem consisting of AI Sweden partners that are suppliers of hardware, software and solutions.

The infrastructure network group today consists of representatives from the National Supercomputer Centre (NSC) at Linköping University, PDC at KTH, Chalmers Center for Computational Science and Engineering (C3SE), Helsingborg city, RISE ICE Data Center, RISE Centre for Applied AI, Statistics Sweden (SCB), Swedish Tax Agency (Skatteverket), Vinnova, Swedish University of Agricultural Sciences (SLU), National Library of Sweden (KB), Gothenburg School of Economics, Wallenberg AI, Autonomous Systems and Software Program (WASP), EuroCC National Competence Centre Sweden (ENCCS) and many more, and is open to further members.

The Infrastructure Expert Group is sharing knowledge relating to a broad range of infrastructure-related topics, building on technical knowledge and learning from the previous experiences of AI Sweden and all the participating partners. The overall idea was initially to discuss things such as:

  • building infrastructure, hardware and peripherals;

  • safety and security policies, cyber and information security;

  • user models and policies for accessing common resources;

  • scheduling, scaling, efficient use of overcapacity;

  • user interfaces;

  • studies of usage and user patterns; and

  • training and learning within ethics and technologies.

During one of the first network meetings, a workshop was set up to identify the most important topics and challenges based on input from the participating member representatives. The outcome of the workshop was a list of areas that, since then, has been guiding the network administration in planning events. The areas include:

  • trends in hardware and software;
  • federated learning and edge computing;
  • scheduling, job handling and large data transfers;
  • security and sensitive data;
  • machine learning operations (MLOps); and
  • current topics of research and use-case studies.

So far, the following topics have been covered:

  • Lessons from building the AI Sweden infrastructure – Asymptotic.AI;

  • Outlook Europe – EuroCC/ENCCS;

  • Security for ML/ML for security – RISE;

  • Local government and AI usage – HPE; and

  • NLP models: Data, requirements and lessons from BerzeLiUs and ENCCS – RISE/AI Sweden & the Swedish Royal Library.

If you work at one of AI Sweden’s partners in a position relevant to any of the topics that are being addressed, please contact Niclas Fock ( niclas.fock@ai.se ) or Fredrik Viksten ( fredrik.viksten@ai.se ) at AI Sweden to lend your expertise to the endeavour.