{"id":27867,"date":"2025-10-02T10:31:40","date_gmt":"2025-10-02T08:31:40","guid":{"rendered":"https:\/\/wp.eurestools.eu\/key_achievement\/ai-federation-framework\/"},"modified":"2025-10-02T10:31:40","modified_gmt":"2025-10-02T08:31:40","slug":"ai-federation-framework","status":"publish","type":"key_achievement","link":"https:\/\/wp.eurestools.eu\/de\/key_achievement\/ai-federation-framework\/","title":{"rendered":"Rahmenwerk der KI-F\u00f6deration"},"content":{"rendered":"","protected":false},"featured_media":0,"template":"","class_list":["post-27867","key_achievement","type-key_achievement","status-publish","hentry"],"acf":{"logo":"https:\/\/smart-networks.europa.eu\/wp-content\/uploads\/2025\/08\/sunrise-6g.png","diagram":"https:\/\/smart-networks.europa.eu\/wp-content\/uploads\/advanced-cf7-upload\/AI_Federation_framework_Architecture2025061403.png","top10":"","project_name":"SUNRISE-6G","ka_number":"3","description":"The SUNRISE-6G AI federation framework was developed for the sharing and collaborative deployment of AI models across multiple testbeds, which is currently hindered by the existing fragmented and proprietary AI\/ML approaches. Currently, most testbeds develop and deploy their own AI models optimised for their specific network setups, often using proprietary datasets and local ML workflows. Additionally, heterogeneity in data formats, APIs, and orchestration frameworks makes it difficult to share or federate models. To address this, SUNRISE-6G promotes the development and sharing of AI\/ML models and workflows across all federated facilities, enabling them to collaborate and share resources and AI models while continually improving their performance. The SUNRISE-6G AI Federation framework leverages the advancements in MLOps, AI Operations (AIOps), and Foundation Models (FM) to facilitate the building of incremental repositories of reusable AI\/ML models and extracting synergistic knowledge from serving models and workflows. The following aspects are addressed in the framework:\n'\u2022 A set of industry standard tools and components, covering the deployment of a global MLOps stack\n'\u2022 Reusable time-series foundation models developed within federation, which are designed to support tasks such as zero-shot learning and automated modulation classification. These models are made available for fine-tuning and domain adaptation via the global MLOps stack, facilitating testbed collaborations.\n'\u2022 AI\/ML experimentation and modelling pipelines implemented at the local testbed level. These pipelines represent key innovations in local MLOps practices and form the foundation for scalable, automated, and testbed-specific AI experimentation workflows within the federated AI framework.","references":"The framework implementation is described in D4.2 (in BSCW) \nSUNRISE-6G contributed to 3GPP SA2 a new solution that enables vertical federated learning across different testbeds, alinged with the AI Federation approach: https:\/\/www.3gpp.org\/ftp\/tsg_sa\/WG2_Arch\/TSGS2_162_Changsha_2024-04\/Docs\/S2-2403921.zip","category":"CAT-1: Significant Technology Development","sub_categories":"AI\/ML;","call":"Call 2"},"_links":{"self":[{"href":"https:\/\/wp.eurestools.eu\/de\/wp-json\/wp\/v2\/key_achievement\/27867","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/wp.eurestools.eu\/de\/wp-json\/wp\/v2\/key_achievement"}],"about":[{"href":"https:\/\/wp.eurestools.eu\/de\/wp-json\/wp\/v2\/types\/key_achievement"}],"wp:attachment":[{"href":"https:\/\/wp.eurestools.eu\/de\/wp-json\/wp\/v2\/media?parent=27867"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}