du, BubbleRAN, and Khalifa University Demonstrate Arabic Voice-Controlled Autonomous 5G Networks

du,  BubbleRAN,  and Khalifa University  showcase an industry and region’s first demonstration of how Arabic voice commands (“intent”) can be seamlessly transformed into validating network actions, executed through O-Radio Access Networks (O-RAN) and AI-Radio Access Networks (AI-RAN) compliant Telco API’s.

The demonstration is a result of a unique industry-academia collaboration, combining reasoning-capable Large Language Models (LLMs) and deep telecom domain expertise – covering local regulation, best practices and user behaviour to deliver actions that are explainable, compliant and production ready.

“As enterprises expand their private 5G footprints, the cost and complexity of network operations remain a critical concern. At du, we see immense potential in O-RAN automation and network intelligence to make private 5G deployments more efficient, scalable, and adaptive to customer-specific use cases. By embracing openness and AI-driven automation, we can deliver both cost efficiency and performance agility across our next-generation networks.”

– Mr. Saleem AlBlooshi, Chief Technology Officer (CTO), du 

The usage of Arabic-Native Telecom LLM enables handling of telecom terms and dialects without relying on English translation, ensuring accuracy. Reflection of local data, regulations, and use cases is guaranteed. The embedded reasoning justifies decisions step-by step by explaining and optimizing network choices transparently in the user’s own language. Operational efficiency is accelerated as engineers manage networks using Arabic voice, speeding automation and reducing errors.

“We are making networks intelligent by letting them speak the language of their users. This is the world’s first step towards autonomous 5G/6G. For the first time, an operator can talk to the network in Arabic, and the network understands, reasons, and acts.”

– Professor Navid Nikaein, Chief Executive Officer, BubbleRAN

The global impact is to enable intelligent automation with zero-touch network management, improving performances and enhancing the user experience in dynamic scenarios offering a culturally aware approach to both operators and engineers.

“The usage of various network optimization tools (TelecomGPT Arabic, specific AI agents and rApps) developed by our research teams and partners is very efficient in this demo. The impact is to enable intelligent automation, accelerating progress toward fully autonomous networks.”

– Professor Merouane Debbah, Director of 6G research center, Khalifa University

Why does it matter?

• Autonomous Operations: from Intent to Action, an Arabic voice command (“intent”) is transformed into the right network configuration or lifecycle operation, then validated and applied via O-RAN rApps.
• Agentic Intelligence: BubbleRAN’s MX-AI framework deploys and coordinates a team of AI agents that automate network operations using LLMs, SLMs, algorithms, and APIs. Task-specific agents are orchestrated by higher-level agents to collaborate and resolve complex network operations.
• Arabic-Native Networks: Showcasing how LLMs can power Arabic-language interaction with telecom networks. The system understands and reasons over Arabic user intents to generate precise network actions, proving that network intelligence can now speak the language of its users.
  Availability

BubbleRAN’s MX-AI solution, published in their latest software release Crimson, is available for deployment on BubbleRAN MX-PDK or other 5G solutions. The Arabic LLM showcased here, is a locally fine-tuned model; the base model is publicly available for research and non-commercial use through Hugging Face.