Concept and Methodology 

6G-DALI aims to develop an innovative data management system, based on Gaia-X and IDS principles, to handle data collected from 6G testbeds and Digital Twins. It also seeks to provide a complete AI framework for 6G, focusing on two main parts: AI experimentation as a service using MLOps and efficient data collection and storage using DataOps. The 6G-DALI DataOps pillar ensures clean, processed data in the 6G Data Lake for training and validating machine learning models within the MLOps pillar. It also focuses on social acceptance of 6G and AI trustworthiness, progressing through the following key phases. 

Phase I defines the 6G-DALI concept, goals, and requirements. It focuses on understanding the business needs, setting up performance measures, and outlining the necessary AI framework for MLOps and DataOps. It also brings together practical examples to help shape the AI framework and determine what is needed for each part of the system. 

Phase II focuses on designing key innovations for 6G-DALI. For DataOps, it includes integrating Gaia-X, enhancing the 6G Dataspace, improving data with AI, and creating algorithms for testbed selection. For MLOps, it works on distributed AI systems, ensuring AI trustworthiness, and adapting testbeds using Digital Twins and other tools for large experiments 

Phase III involves testing and assessing the innovations from Phase II to make sure they work effectively within the AI framework. Different metrics will be used to evaluate specific issues (e.g., Time to detect ML model drifting, fidelity and actionability, etc The findings will then be used to refine and improve the solutions from Phases I and II. 

Phase IV aims to validate the 6G-DALI framework through PoC experiments, integrating all components to support DataOps and MLOps. It ensures the solutions meet the project’s goals, with iterative feedback from Phase III to fine-tune and achieve the expected KPIs and KVIs.