catalysts logo

Gain access to resources and project updates

Register or log in to save your details for future use
First name
Last name
Business email
Company name
Job title

TM Forum will be processing the above information, with the assistance of our service providers located within and outside the European Union, to manage your registration to this event or report download, as well as to keep you informed about our services and products, future events and special offers, the organization of events, providing training and certification, and facilitating collaboration programs. Privacy policy

I wish to receive further information from the Catalyst Team about their products and service by electronic means. Check the "Team Members" section of this Catalyst Project to review the companies that will receive your information. Companies may join the project in the future, so please check back periodically for any updates

logologo
All projects

GENAR: GenAI-enabled anomaly detection for RAN

URN C25.0.825
Topics 5G, AI (Artificial Intelligence), Autonomous networks

Transforming telecom networks with generative AI accelerated RAN anomaly detection to boost network performance, streamline operations, and ensure exceptional service.

featured image
Traditional AI tools struggle to keep up with the nuances of detecting anomalies in RAN networks and its evolving demands. They often miss subtle relationships in data, generate false positives, and fail to detect new types of anomalies. This Catalyst introduces advanced anomaly detection framework powered by generative AI (genAI), time-series analytics, and continuous model learning. The solution builds on Capgemini’s EIRA framework and integrates Amazon SageMaker’s MLOps for automated training and inference. It uses spatial-temporal event correlation to analyze real-time KPI data such as latency, throughput, and reliability. Additionally, Amazon Bedrock adds genAI capabilities, simulating failure scenarios and offering rich contextual insights through time-series foundation models. To support scale and performance, the architecture includes Amazon Glue for live data integration and S3 for cloud-native storage. Moreover, it aligns with TM Forum’s CLADRA framework, enabling closed-loop detection, resolution, and dynamic network optimization without manual intervention. These features ensure faster incident handling and increased automation maturity. Unlike traditional systems, GENAR can process both structured and unstructured data—including logs and alarms. It also adapts continuously, improving model accuracy and responsiveness as conditions evolve. As a result, CSPs gain greater visibility, reduced mean time to detect and resolve issues, and better customer experience. The team will track performance using key KPIs such as 95% detection accuracy, 80% early anomaly mitigation, and a 30% reduction in operational costs. Furthermore, the platform’s open API model ensures compatibility with multi-vendor environments, while supporting extensions into IoT, edge, and compliance use cases. By blending genAI, Graph ML, and TM Forum best practices, this Catalyst provides a proactive, scalable, and intelligent anomaly detection system. With this solution, CSPs may have the assurance that operational complexity can finally be turned into competitive advantage.

Team members

Aira Technologies Inc logo
Amazon Web Services, Inc. logo
BT Group plc logo
Champion
Capgemini logo
Deutsche Telekom AG logo
Champion
TELEFONICA logo
Champion
TELUS logo
Champion

Related projects