Accelerating open and AI-native RAN with BubbleRAN MX-AI

This talk introduces BubbleRAN MX-AI, an open, modular platform and development toolkit designed to accelerate the design, deployment, and orchestration of multiple AI agents and rApps, aligned with O-RAN, AI-for-RAN, and AI-on-RAN initiatives.



Three foundational pillars of MX-AI will be presented and exemplified: (1) Open ecosystem, providing standards-aligned O-RAN interfaces with support for xApps/rApps/agents, interoperable RU/DU/CU stacks, and pluggable RICs; (2) Open platform enabling reproducible pipelines for data collection, feature engineering, and MLOps (training, evaluation, and CI/CD for models), plus hardware-in-the-loop and digital-twin environments to de-risk field trials; (3) Open benchmarking with curated datasets, KPIs, and reference scenarios (scheduling, slicing, energy, and anomaly detection) to enable apples-to-apples comparisons and transparent progress.