When MLOps meets NWDAF to enable autonomous next generation network analytics

The shift toward 6G intensifies the need for Zero-Touch Service Management (ZSM) to enable autonomous network operations.  Central to this vision is the Network Data Analytics Function (NWDAF), a standardized entity defined by the 3rd Generation Partnership Project (3GPP) to deliver data-driven intelligence within the 5G core network. While 3GPP Release 17 introduced the architectural separation of NWDAF into the Analytics Logical Function (AnLF) and the Model Training Logical Function (MTLF), the specifications leave the model lifecycle management largely undefined.  This creates a significant specification-to-implementation gap that hinders truly autonomous analytics. Furthermore, existing NWDAF open-source initiatives lack coherent ML Operations (MLOps) integration, particularly for automated model training and lifecycle orchestration. 
This paper addresses these limitations by presenting the first comprehensive, standards-aligned open-source NWDAF implementation fully compliant with 3GPP Release 17. Our contribution extends the OpenAirInterface NWDAF by establishing the AnLF–MTLF architectural separation with an MLOps-enabled MTLF that autonomously orchestrates the model lifecycle. The proposed NWDAF is validated across representative 3GPP use cases, demonstrating stable autonomous model training, efficient end-to-end analytics provisioning, and consistent predictive performance with low inference latency.