End-to-end AI lifecycle management for 6G: Bridging DataOps and MLOps

As 6G networks evolve toward a data-centric architecture with full automation capabilities, the softwarization of network functions enables access to a wide spectrum of data across technical, management, and operational domains. This rich data provides a strong foundation for building efficient, dynamic, and scalable network environments, where Artificial Intelligence (AI) and Machine Learning (ML) play a vital role in enabling data-driven intelligence within 6G networks. Consequently, streamlining AI/ML workflows becomes essential to fully harness the potential of 6G. Given the massive volume of data generated and the extensive applicability of AI/ML across various layers of the 6G architecture, including the Radio Access Network (RAN), Core, and Edge, managing AI/ML workflows at scale presents significant challenges. To address these challenges, we propose a novel microservice-oriented DataOps-MLOps framework designed to streamline AI/ML workflows across all domains. The proposed framework natively supports model training, model versioning and incorporates an automated data pipelining approach that facilitates seamless data collection and real-time inference. To validate the framework’s effectiveness, we implemented a traffic anomaly detection use case involving end-to-end (E2E) ML model training, deployment, and real-time inference. Experimental results demonstrate that the proposed framework significantly streamlines AI/ML operations, unlocking the full potential of AI/ML integration in future 6G networks.