Chenhua Shi

This entry showcases the development of an autonomous AI solution that is transforming how telecommunications networks are operated and maintained. By combining Agentic AI, Multi-Agent Systems (MAS), fine-tuned Small Language Models (SLMs), causal AI, and LLM-powered rApps, the solution automates complex troubleshooting tasks across both Radio Access Networks (RAN) and Packet Core, enabling faster, more accurate, and more reliable network operations.

Unlike conventional AI assistants, the system orchestrates specialized AI agents that collaboratively perform planning, data retrieval, root cause analysis (RCA), remedy recommendation, and execution, delivering explainable recommendations that network engineers can trust. To further improve performance in highly specialized telecom environments, the project introduced a novel domain-grounded synthetic data generation framework for fine-tuning telecom-specific language models, creating compact AI models optimized for production-scale deployment.

The innovation achieved a six-fold reduction in network troubleshooting time and a 10% improvement in diagnostic accuracy, significantly improving operational efficiency, service reliability, and customer experience. Beyond individual AI models, the project also co-developed the ELIA platform, enabling rapid deployment of AI-powered rApps and accelerating enterprise adoption of autonomous network operations.

The solution has progressed from research to real-world implementation through collaborations with leading telecommunications operators and technology partners. Its technical contributions have been recognized through peer-reviewed IEEE publications, international conference presentations, and Ericsson's Top Performance Award, demonstrating both technical excellence and measurable industry impact. This work represents an important step toward trustworthy, scalable, and autonomous AI for next-generation telecommunications networks.