When MLOps meets NWDAF release 17/18

This demonstration showcases how scalable Machine Learning Operations (MLOps) platforms can enhance and extend the 3GPP Network Data Analytics Function (NWDAF) as defined in Release 18. The 3GPP specification TS 23.288 defines the high-level functional architecture of the NWDAF, introducing the Model Training Logical Function (MTLF) and the Analytics Logical Function (AnLF). However, it explicitly excludes the details of model lifecycle management, such as model storage, deployment, and monitoring, leaving these aspects out of scope. In parallel, 3GPP TS 28.105 specifies a general framework for model management within the Operations and Maintenance (OAM) domain. While it provides overarching guidance for model operations, TS 28.105 is not specifically tailored to the NWDAF context and lacks concrete implementation details to ensure scalability, reliability, and integration with service-based 5G Core functions. This overlapping yet incomplete coverage between TS 23.288 and TS 28.105 creates a critical implementation gap for operators aiming to deploy production-grade analytics and machine learning functions within NWDAF. Without a clear alignment between architectural definitions and operational frameworks, the realization of intelligent, automated network analytics remains limited. To address this gap, our demonstration integrates an open-source MLOps stack consisting of Kubernetes, MLflow, and MinIO with the NWDAF architecture. This integration introduces robust capabilities for model registry, metadata management, artifact storage, and continuous monitoring. Our implementation enhances the OpenAirInterface (OAI) NWDAF (Release 15 baseline) to align with Release 18 specifications, offering standardized APIs for both MTLF and AnLF functions. The resulting solution provides a unified and scalable model management framework that bridges 3GPP architecture and OAM practices. It enables service-based, automated, and resilient machine learning workflows for 5G Core intelligence and establishes a practical foundation for intelligent, self-optimizing networks in the evolution toward 6G.