D7.4 Service Portfolio

LUMI AI Factory Service Center Empowering Europe’s AI Ecosystem D7.4 Service Portfolio 2 D7.4 Service Portfolio D7.4 Service Portfolio 3 Project Title LUMI AI Factory Service Center Project Acronym LUMI-AIF Project Number 101234208 Type of Action HORIZON-JU-RIA Topic HORIZON-JU-EUROHPC-2025-AI-01-IBA-01 Starting Date of Project 01.03.2025 Ending Date of Project 29.02.2028 Duration of the Project 36 months Website lumi-ai-factory.eu Work Package WP7 Task T7.3 Service portfolio development Lead Authors Roger Kvam (Sigma2) Contributors Basir Sedighi (Sigma2), Kalle Huhtala (CSC) Peer Reviewers Martin Duda (IT4I), Aleksi Kallio (CSC) Version 1.0 Due Date 29.08.2025 Submission Date 29.08.2025 Dissemination level X PU: Public SEN: Sensitive – limited under the conditions of the Grant Agreement EU-RES. Classified Information: RESTREINT UE (Commission Decision 2005/444/EC) EU-CON. Classified Information: CONFIDENTIEL UE (Commission Decision 2005/444/EC) EU-SEC. Classified Information: SECRET UE (Commission Decision 2005/444/EC) D7.4 Service Portfolio 4 Version History Revision Date Editors Comments 0.1 27.05.2025 Roger Kvam First version 0.2 28.05.2025 Roger Kvam Updated with plan for development of roadmap 0.3 23.06.2025 Roger Kvam Changed structure to align with document for customer-oriented representation 0.4 24.06.2025 Basir Sedighi & Roger Kvam Added text to structure 0.5 15.08.2025 Roger Kvam Rewriting first section 0.6 24.08.2025 Roger Kvam Kalle Huhtala Extending the executive summary, adding items to the Glossary, completing 3.4.1 and replacing chapter 6 with reference to the Wiki. Adding the chapter on Service Catalogue. 0.7 29.8.2025 Kalle Huhtala Final editing 1.0 29.8.2025 Rebekka Lampola Quality check performed by the PMO and sent to official review Glossary of terms Item Description AI Artificial Intelligence: The theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages. AIF AI Factory: A centralized hub providing computing resources, data, tools, and expertise to develop, deploy, and scale AI applications. Airflow An open-source workflow orchestration platform originally developed by Airbnb, designed to author, schedule, and monitor data and machine learning pipelines. API Application Programming Interface: A set of rules and protocols that allows different software applications to communicate with each other. Batch Job Scheduling A method of managing and executing tasks (jobs) on a computing system without direct user interaction, typically used in High-Performance Computing to optimize resource usage. CSC The Finnish IT center for science, which hosts and operates the LUMI supercomputer on behalf of the EuroHPC JU. D7.4 Service Portfolio 5 CUDA Compute Unified Device Architecture: A parallel computing platform and programming model created by NVIDIA for general computing on its graphical processing units (GPUs). CV Computer Vision: A field of AI that trains computers to interpret and understand the visual world from digital images or videos. Dataset A structured collection of related data, typically presented in a tabular, textual, visual, or numerical format, organized for analysis, training of AI/ML models, or other computational use. Datasets may contain raw or curated information and can vary in size from small samples to petabyte-scale collections. In the context of the LUMI AI Factory, datasets are provided as “Dataset-as-a-Service,” giving users standardized, secure, and often pre-processed data resources for research, innovation, and AI development, including sensitive and confidential data where compliance with regulations is required. Digital Twin A virtual model designed to accurately reflect a physical object, process, or system. It is used for simulation, testing, and monitoring. EuroHPC JU European High-Performance Computing Joint Undertaking: A legal and funding entity that enables the European Union and participating countries to coordinate their efforts and pool their resources to deploy world-class supercomputers and technologies. Federated Learning A machine learning technique that trains an algorithm across multiple decentralized devices or servers holding local data samples, without exchanging the data itself. Fine-tuning The process of taking a pre-trained model (like a large language model) and further training it on a smaller, specific dataset to adapt it for a particular task. Foundation Model A large-scale AI model trained on a vast quantity of broad data that can be adapted ("fine-tuned") to a wide range of downstream tasks. GenAI Generative Artificial Intelligence: A type of AI that can create new and original content, such as text, images, music, or code. GPU Graphics Processing Unit: Specialized electronic circuits designed to rapidly manipulate and alter memory to accelerate the creation of images in a frame buffer intended for output to a display device; their parallel structure makes them ideal for complex AI and machine learning tasks. HIP Heterogeneous-compute Interface for Portability: A C++ runtime API and kernel language that allows developers to create portable applications for AMD and NVIDIA GPUs from a single source code. HPC High-Performance Computing: The use of supercomputers and parallel processing techniques for solving complex computational problems. IT4I IT4Innovations National Supercomputing Center at VŠB – Technical University of Ostrava, Czechia. A key national research infrastructure and a member of the LUMIAIF consortium. KAM Key Account Manager: An individual responsible for managing and nurturing relationships with an organization's most important customers. Kubeflow An open-source platform for developing, orchestrating, and deploying scalable machine learning workflows on Kubernetes KPI Key Performance Indicator: A quantifiable measure of performance over time for a specific objective. KPIs provide targets for teams to shoot for, milestones to gauge progress, and insights that help people across the organization make better decisions. LLM Large Language Model: An advanced type of AI model designed to understand, generate, and process human language, trained on massive datasets of text and code. Examples include Llama and GPT models. M Month, typically followed by a number denoting the sequence from the project start D7.4 Service Portfolio 6 MFA Multi-Factor Authentication: A security process that requires users to provide two or more verification factors to gain access to a resource, such as an application or online account. MLOps Machine Learning Operations: A set of practices that aims to deploy and maintain machine learning models in production reliably and efficiently. It combines machine learning, data engineering, and DevOps. MLFlow An open-source platform for managing the end-to-end machine learning lifecycle. MOOC Massive Open Online Course: A free online course available for many people to enrol. MS Milestone, typically followed by a number identifying it. Multitenancy A software architecture where a single instance of a software application serves multiple customers (tenants), with each tenant's data isolated and remaining invisible to other tenants. NLP Natural Language Processing: A subfield of AI focused on enabling computers to understand, interpret, and generate human language. QPU Quantum Processing Unit: The core processor of a quantum computer, which uses the principles of quantum mechanics to perform calculations. Sandbox In an AI/software context, an isolated testing environment that enables users to run programs or open files without affecting the application, system, or platform on which they run. Service Catalogue A customer-facing list of all live services offered along with relevant information about these services. It can be regarded as a filtered, user-friendly view of the Service Portfolio. Service Portfolio An internal list that details all the services offered by a service provider, including those in preparation, live, and discontinued. It contains technical specifications, costs, risks, and value propositions. SME Small and Medium-sized Enterprise: A business whose personnel numbers and revenue fall below certain limits. Trustworthy AI An approach to artificial intelligence that emphasizes the importance of developing AI systems that are lawful, ethical, and technically robust, ensuring they respect human values and fundamental rights. Try-n-buy A service model that allows organizations to experiment with computing and data resources on a limited scale before committing to full adoption. WP Work Package: A major sub-division of a project. Each work package has a set of defined tasks, deliverables, and objectives. D7.4 Service Portfolio 7 Executive summary The LUMI AI Factory is a three-year European initiative from March 2025 to February 2028 designed to accelerate AI innovation by providing a comprehensive one-stop-shop built on the LUMI supercomputer. It aims to empower SMEs and start-ups, researchers and public sector across Europe. This document will describe the roadmap of the core services of the LUMI AI Factory and the planned roadmap for delivery of the maturity levels of the services, as well as when the services will be deployed, and to whom the services will be available. The Service Portfolio will be divided into a Service Catalogue for external use and a Service Portfolio for internal use. The document includes a Service Catalogue draft with initial information on the availability of the services. Key services will be progressively rolled out and include: • Powerful AI Computing: Scalable access to GPU-accelerated resources for training and deploying everything from conventional ML models to massive foundation models, including secure environment for sensitive data • Data Services: Access to a curated library of datasets ("Dataset-as-a-Service") and expert support for data management and integration • Expert Support & AI Adoption: Direct consultation for technical challenges, grant applications, and business-focused programs like "Try & Buy" projects to ease into large-scale AI • Training & Skill Development: A full range of courses, hackathons, summer schools, and an accelerator program for AI start-ups • Collaboration Hubs: A physical co-working hub in Otaniemi, Finland, including virtual spaces, matchmaking and community building across borders Through this roadmap, the LUMI AI Factory will not only deliver advanced technical services but also measurable impact: strengthening European sovereignty in AI, lowering entry barriers for SMEs, accelerating public-sector digitalisation, and equipping a new generation of AI professionals with the skills to thrive in a rapidly evolving ecosystem. This document will be revised annually twice to reflect in increasing detail the current and evolving status of the service development roadmap, portfolio and catalogue. D7.4 Service Portfolio 8 Table of Contents 1. Introduction............................................................................................................. 10 2. Implementation roadmap and service maturity timeline ........................................ 11 3. Service Portfolio roadmap ........................................................................................ 14 3.1 Core AI computing services 14 3.1.1 Key features 14 3.1.2 Initial setup 14 3.1.3 Roadmap 15 3.2 Access to data 15 3.2.1 Key features 15 3.2.2 Initial setup 15 3.2.3 Roadmap 15 3.3 Support for AI methods 16 3.3.1 Key features 16 3.4 AI adoption for companies 17 3.4.1 Key account manager services in complex projects 17 3.4.2 HPC/AI starter pack 17 3.5 Training courses and skills development 17 3.5.1 Training courses on fundamentals of AI 17 3.5.2 Training courses on large scale AI and High-Performance Computing (HPC) 18 3.5.3 Access to online training materials and MOOCs 18 3.5.4 Enrolment in Summer school or Hackathon 18 3.5.5 Enrolment in Accelerator Program 19 3.6 Co-working spaces 19 3.6.1 Otaniemi AI Hub (Espoo, Finland) 19 3.6.2 Location and access 20 3.6.3 Premises for workshops 20 3.6.4 Personal and team working space 20 3.6.5 Virtual co-working space 20 D7.4 Service Portfolio 9 3.6.6 AI Campus services for students 20 3.6.6.1 Campus locations and activities 20 3.6.6.2 LUMI-AI co-laboratory 20 3.6.6.3 General Information and access 21 3.7 Events and collaboration 21 3.7.1 Enrolling in upcoming events within AI Factory ecosystems 21 3.7.2 Requesting collaborator suggestions from the AI Factory ecosystems 22 3.7.3 Requesting LUMI AI Factory staff to give presentations at your event 22 4. Service Portfolio development roadmap ................................................................. 23 4.1 Inputs 23 4.2 Timelines 23 5. Service Catalogue shown to end users ...................................................................... 23 5.1 Service Catalogue management and data model 24 5.2 Service Catalogue draft 25 6. Deliverables and KPIs .............................................................................................. 28 D7.4 Service Portfolio 16 • M02–M10: Identification of relevant datasets and license negotiation • M14: At least 100 datasets available via the service • M18+: Support for public-private data collaboration, commercial license handling, and automatic data staging 3.3 Support for AI methods 3.3.1 Key features • HPC optimization and AI development ▪ The LUMI AI Factory provides support for optimizing High-Performance Computing (HPC) for AI development. This service is part of the in-depth AI expert guidance and support available to facilitate the successful implementation of AI projects, covering advanced topics in HPC optimization for AI development. ▪ Duration: 2 months • Distributed AI model training ▪ Expert support is offered for distributed AI model training. This falls under the indepth AI expert support provided by the AI Factory and is part of the "AI expert support for methods and code optimization and scalability improvement." The support covers advanced topics including distributed AI model training, model fine-tuning, and hyperparameter search. ▪ Duration 2 month • AI method Consultation Assignment o The LUMI AI Factory offers AIF (AI Factory) specialist support for AI methods. This is provided as in-depth AI expert support to help customers with specific AI projects. The AI methods support team includes specialists for various AI subfields, enabling them to provide customer-specific AI consultation on relevant areas such as: ▪ Natural Language Processing (NLP) and Large Language Models (LLMs) ▪ Computer Vision (CV) ▪ Reinforcement Learning (RL) and Robotics ▪ Digital Twins and AI for Science applications o Duration: 1 month • Request for Customer-specific AI Consultation ▪ Customer-specific AI consultation is a service available through the LUMI AI Factory. This is part of the broad range of AI expert guidance and support offered to facilitate the successful implementation of AI projects. Specialist teams provide this customer-specific AI consultation across various subfields of AI. ▪ Duration: 6 months D7.4 Service Portfolio 17 • Grant application support consultation ▪ Consultation meeting with AI Factory experts, providing support for writing grant applications, such as applications for national innovation funding or European programs including Digital Europe Program or Horizon Europe. Consultation focuses on technical feasibility of the planned AI or HPC approaches, as well as their implementation on the LUMI AI Factory environment. 3.4 AI adoption for companies 3.4.1 Key account manager services in complex projects • Customer Needs Analysis will be conducted to understand the customer's expertise and AI readiness • Guidance to appropriate services • Single point of contact 3.4.2 HPC/AI starter pack Dedicated offering to bring manufacturing, engineering, life sciences and pharmaceutical companies closer to HPC and AI solutions and methodologies through targeted services and support via HPC/AI starter pack for companies, Try & Buy test projects, or consultation for Trustworthy AI. 3.5 Training courses and skills development The LUMI AI Factory places a significant emphasis on training courses and skills development to empower Europe’s AI ecosystem, address the talent gap, and accelerate the adoption and development of AI solutions. This comprehensive approach is designed to benefit a diverse range of stakeholders across Europe. 3.5.1 Training courses on fundamentals of AI The LUMI AI Factory provides foundational learning opportunities through a modular and structured training programme, particularly for users with limited or no AI and/or HPC background. • ”Fundamentals of machine learning”: Covers essential concepts of machine learning, including statistical, probabilistic, and computational principles. • ”Introduction to using AI models and tools”: Provides basic skills for using existing AI models, such as large language models, as part of research and development activities. D7.4 Service Portfolio 18 • ”Practical deep learning”: Delves into deep learning and common neural network architectures, GPU computing, and tools for training and applying deep neural networks in areas like NLP and computer vision. These core courses are scheduled at least twice a year and can be delivered online or onsite. 3.5.2 Training courses on large scale AI and High-Performance Computing (HPC) Specialized training is offered for users with existing cloud or AI expertise to facilitate the continuous updating of their skills. • ”AI for HPC”: Covers AI frameworks (PyTorch), containers, visualization, distributing and scaling AI workloads across multiple GPUs/nodes, and optimization techniques. • ”AI workflows and MLOps”: Focuses on AI engineering techniques for building model training workflows and MLOps to deploy and maintain models, ensuring scalability, reproducibility, and reliability. • ”GPU programming”: Covers low-level GPU programming using technologies like CUDA/HIP, OpenMP/OpenACC, SYCL, and Kokkos for optimal resource utilization. • ”Application performance analysis and optimization”: Involves training on profilers, message passing analysis, hardware counter tools, and optimization techniques. 3.5.3 Access to online training materials and MOOCs High-quality, comprehensive self-learning materials are provided to enable users to acquire skills at their own pace. • MOOCs: – ”LUMI AI Factory computing environment”: Covers the basics of the LUMI and LUMI-AI supercomputer environments. – ”Elements of supercomputing”: Introduces the basic principles of highperformance computing. • Online Documentation: Comprehensive user guides and documentation for software packages and tools. • AI Assistant: A chatbot will be available to answer questions, generate scripts, and analyse data, built on an open-source LLM trained on LUMI-AI. • Integrated service portal: A single point of entry to all information and services, integrating web pages, documentation, and user portals for resource management. 3.5.4 Enrolment in Summer school or Hackathon D7.4 Service Portfolio 19 The LUMI AI Factory actively organizes workshops, hackathons, and summer schools as part of its structured training program. • Hackathons: – “Moving and optimizing your AI jobs to LUMI-AI supercomputer”: Helps users port, modify, and optimize their AI software for the supercomputer. – “AI students hackathon”: A collaborative event based on industry-provided use cases and datasets. • Summer Schools: – “Training and using AI on the LUMI-AI supercomputer”: Covers the LUMI/LUMI-AI environment, AI frameworks, distributed AI, and optimization. – “AI for science summer school”: Integrates computationally intensive simulations with AI and covers the development of digital twins. Participants receive a certificate with a recommendation on corresponding ECTS credits. 3.5.5 Enrolment in Accelerator Program An accelerator program for AI startups is planned to promote the excellence of LUMI-AI and support the development of advanced models. • Launch: The AI startup incubator programme is expected to be launched by Month 10 (approx. December 2025). • Scope: The program will support one batch of 10–20 startups from EuroHPC member states. • Fast-Track Model: A “Try & Buy test project” model allows for free-of-charge testing with a small resource allocation and expert support before regular use. 3.6 Co-working spaces The LUMI AI Factory places a strong emphasis on fostering a collaborative and innovative AI ecosystem through its diverse range of co-working spaces and services. This includes both physical and virtual environments designed to support students, startups, SMEs, researchers, and AI professionals across Europe. 3.6.1 Otaniemi AI Hub (Espoo, Finland) The Otaniemi AI Hub serves as the main physical co-working space for the LUMI AI Factory, strategically located to concentrate AI talent and expertise. D7.4 Service Portfolio 20 3.6.2 Location and access The hub is situated on the Otaniemi campus of Aalto University at Tietotie 2, Espoo, Finland. It is co-located with the upcoming ELLIS Institute Finland and surrounded by key national AI competence centres, startup campuses, and incubators. The area has excellent connectivity to Helsinki international airport and is easily accessible via public transportation. Onboarding activities and tours will be available through the integrated service portal. 3.6.3 Premises for workshops The hub is equipped with advanced physical training facilities well-suited for workshops. These include a fully equipped computer classroom, wireless network access, and a nearby auditorium for larger events. Additional facilities are provided by CSC in Keilaranta and collaborators at Aalto University and the University of Helsinki, allowing for a high number of parallel training sessions. 3.6.4 Personal and team working space The main co-working space features an open office with desks for flexible working. For confidential conversations or focused work, private phone booths are available. The space is administrated by a local coordinator and has permanent on-site staff to assist visitors. 3.6.5 Virtual co-working space The LUMI AI Factory complements its physical spaces with a robust virtual working space to enable effective collaboration across participating countries. This digital environment provides a suite of tools for team collaboration, including: • Shared document management • Co-creation tools • Work management tools • Discussion forums • Videoconferencing tools These resources are continuously available to support collaboration and innovation within the AI ecosystem for the duration of the project. 3.6.6 AI Campus services for students 3.6.6.1 Campus locations and activities The AI Factory Hub in Otaniemi will include a dedicated student hosting facility. Smaller AI Campuses are also planned for Ostrava (hosted by IT4I) and other partner facilities like the University of Helsinki’s Kumpula campus. These campuses offer activities such as study groups, peer-tutoring, matchmaking events, networking, visits, and lectures. Students will also receive user accounts with small computing environments for practical HPC familiarization. 3.6.6.2 LUMI-AI co-laboratory D7.4 Service Portfolio 21 This key service offers user-friendly web-based access to short-time GPU computing resources, ideal for course exercises, self-study, and thesis work. It provides Jupyter Notebooks and other web-based interfaces, allowing access to training environments and materials anytime without needing local software installations. 3.6.6.3 General Information and access The LUMI AI Factory offers an integrated service portal which serves as a single point of entry to all information and services. This portal will integrate public webpages, user documentation, and user portals for managing resources and directly using services via Open OnDemand. Open services will be available to all participants from EuroHPC member states via this portal. For tailored services, a Customer Needs Analysis may be conducted to guide users to suitable offerings. 3.7 Events and collaboration The LUMI AI Factory is designed to be a comprehensive hub for AI innovation, offering various services to foster collaboration, skill development, and the widespread adoption of AI across Europe. This includes both opportunities to participate in events and to leverage the expertise and networks of the AI Factory. 3.7.1 Enrolling in upcoming events within AI Factory ecosystems • General access to event information: The primary entry point for all information and services, including event schedules and registration, is the integrated service portal of the LUMI AI Factory. • Types of events: A range of activities are designed for various skill levels and target groups, including: – Structured training programs – Workshops on specific domains – Hackathons based on industry use cases – Summer schools – Webinars and lectures – Matchmaking and networking events • European-wide events: Joint training events, workshops, and hackathons are planned with other EuroHPC AI Factories (expected two per year). The AI Factory also supports an Annual European AI Factories event to connect the entire AI community. • National ecosystem events: The AI Factory actively engages with and supports national AI ecosystems in consortium countries (Czechia, Denmark, Estonia, D7.4 Service Portfolio 22 Finland, Norway, Poland). National partners will contribute to the training portfolio, with events accessible through the common portal. • Enrolment process: The integrated service portal will facilitate course and event registration. Most open services are available to all participants from EuroHPC member states. 3.7.2 Requesting collaborator suggestions from the AI Factory ecosystems The LUMI AI Factory is a catalyst for collaboration and innovation within the AI ecosystem. • Matchmaking services: A core component is matchmaking between AI developers, data providers, and users to facilitate new partnerships, especially for public-private collaborations and grant proposals. • Access to ecosystem insights: The AI Factory’s extensive network provides deep insight into relevant AI ecosystems, helping users identify and connect with suitable collaborators. • Stakeholder engagement: Key Account Managers assist in defining project scope and encourage companies to build collaborative projects with universities, research institutes, and public bodies. • AI Campus and incubator programs: The AI Campus facilities host matchmaking events. The AI startup incubator program (launching approx. Dec 2025) provides further avenues for connecting with promising startups. 3.7.3 Requesting LUMI AI Factory staff to give presentations at your event The LUMI AI Factory places a strong emphasis on communication, outreach, and knowledge dissemination. • Expert presentations: The AI Factory’s diverse pool of AI, HPC, and domain-specific experts are available to give presentations on a variety of topics. • Communication and outreach strategy: A dedicated work package (WP8) is tasked with raising awareness of the AI Factory’s resources and supporting their uptake through efficient communication and dissemination. • Knowledge sharing and training: The AI Factory’s experts are actively involved in educating the community through training events, webinars, workshops, and hackathons. • How to request: To initiate a request for collaborator suggestions or for a presentation, users can contact the relevant teams (e.g., Customer Engagement, AI Expert Support, or Communication) through the integrated service portal. D7.4 Service Portfolio 23 4. Service Portfolio development roadmap 4.1 Inputs • Initial customer survey developed and performed through task 7.1 and Deliverable 7.1 o Needs of industry o Needs of SMEs and startups o Needs of public administration o Needs of academia and research • Input from continuous screening in task 7.1 of potential: o Customers and Users o Technologies and latest innovation o Competitors and other market data • Business model planning, cost versus reward analytics in task 7.2 • Customer process o Customer journey analytics from onboarding processes in task 2.1 and task 2.2 (D2.2) o Customer needs analytics in task 2.3 ▪ In-depth discussions with customers ▪ Well-structured surveys ▪ Maturity surveys prior to onboarding new customers (D2.3) o HPC and AI project implementation experience o Dialog with KAM 4.2 Timelines • Review roadmap in M6, M12, M18, M24, M30 5. Service Catalogue shown to end users The Service Catalogue is defined as follows: “Customer-facing list of all live services offered along with relevant information about these services. Note 1: A Service Catalogue can be regarded as a filtered version of and customers’ view on the Service Portfolio.” Some services are already available with existing LUMI supercomputer. Others can be made available to customers soon, while some services need the actual LUMI-AI and finally LUMI-IQ machines to be available. D7.4 Service Portfolio 24 5.1 Service Catalogue management and data model The information presented in the Service Catalogue must be user-friendly and informative, considering the expected spectrum of users ranging from startups and traditional enterprises to academia and research professionals. Producing suitable service descriptions calls for co-operation between Service Owners and Communications. A repository (table, database, existing service, to be defined) for storing service information is needed, along with the procedures for managing this information. A data model / template for describing the service details is needed. Such a template will be defined ASAP for the purpose of opening the LUMI AIF website in early September 2025. The service description template should include e.g. the following topics that are useful to end users. For managing services internally, many more details would be required (e.g. service description fields in the Service Now -database used by CSC to maintain its service offering): • Service Name: [Enter AI Service Name Here] • Service Summary: [Brief overview of the AI service] • Service Category: [e.g., Machine Learning, NLP, Computer Vision, Training, Consultation…] • Service Owner: [Team or individual responsible] • Target Users: [Who can use this service] • Service feature list • [Feature 1] • [Feature 2] • Service Levels (SLAs): [Availability, response time, etc.] • Request Process: [How to request the service] • Support Information: [Contact details, support hours] • Cost and Billing: [Pricing model if applicable] • Dependencies: [Other services or systems required] • Security and Compliance: [Data protection and regulations] • Change and Maintenance Schedule: [Planned updates] • Service Status: [Current operational status] • etc. D7.4 Service Portfolio 25 5.2 Service Catalogue draft The following list is the first draft of LUMI AI Factory Service Catalogue. The services listed here will be shown with appropriate information on the LUMI AIF website, with the services marked either as available or with a note of estimated availability. The eight categories of services correspond to the Service Portfolio. An estimate for the start of service availability is marked in parentheses after the service name. E.g. “2026” means the service will be available in that year, not necessarily beginning that year. 1. AI computing capacity 1.1. Small computing package (Availability: 2025+) • Suitable for training conventional machine learning models 1.2. Medium computing package (2025 with limitations. Fully in 2026) • Suitable for fine-tuning large language models 1.3. Large computing package (2025 with limitations. Fully in 2026) • Suitable for training large language models of up to 10B parameters 1.4. Grand challenge package (2025 with limitations. Fully in 2027) • Suitable for training large language models of 10-500B million parameters or other very large foundation models 1.5. Quantum computing package (Not yet) • Suitable for computing tasks requiring the combining of quantum computing and AI or HPC computing 2. Access to data • Access to Datasets-as-a-Service (DaaS) (2025: Partially available. Fully 2026+) • Data catalogue (DaaS component) (2025: Partially available. Fully 2026+) • User support (DaaS component) (2025) • Data permit management (DaaS component) (2026+) • Data transfer (DaaS component) (2026+) • LUMI AIF data environment user guide (2025 partially available, 2026+) • Storage contract (2025 piloting, 2026+) • Data space integration (2026+) • Data streaming (2025 piloting, 2026+)