Mavenir introduced an Operations Co-Pilot that uses generative AI (GenAI) to improve service-level agreements (SLAs) and optimize operational efficiency for communications service providers (CSPs).
Mavenir’s Operations Co-Pilot for RAN Service Assurance (RSA), which was developed in collaboration with NVIDIA and AWS, automates network troubleshooting, enabling mobile network operators to rapidly and accurately anticipate network issues and address detected network impairments before they become critical. The Operations Co-Pilot significantly reduces manual debugging effort, development and maintenance lead times to enhance IT operations, service availability and delivery for increasingly complex mobile networks.
Utilizing the power of GenAI with large language models (LLMs) trained on detailed key performance indicators (KPIs), counters, logs and traces from network infrastructure, Mavenir, together with NVIDIA and AWS, is creating a suite of Operations Co-Pilot solutions that will transform the operations management of telecom networks through intelligent automation. This new framework – built on NVIDIA Tensor Core GPUs and the NVIDIA AI Enterprise software platform for generative AI running on AWS – leverages Mavenir’s Open RAN architecture to provide accessibility into multiple open interfaces that can deliver the data needed to train and optimize domain-specific telecom LLMs.
Mavenir’s Chief Technology and Strategy Officer Bejoy Pankajakshan commented: “Our new Operations Co-Pilot framework has the potential to be a game-changer for operators, delivering a wealth of fault prediction and root cause analysis capabilities with AI-powered accuracy and speed. This industry-first solution is built on the transformative foundation of NVIDIA GPUs operating on AWS and enabled by Mavenir data, which can be accessed via numerous interfaces in the Open RAN and in the Packet Core for complete end-to-end visibility.”
He added: “The three key use cases of Core Dump Analysis, Log Similarity Search feature and Log Anomaly Detection can be used together to maximize the automated efficiency gains. Combining information from these trained LLMs will provide broader and deeper coverage of all fault and anomaly scenarios, enabling very early prediction of potential problems in the network. Moreover, as these solutions evolve, the models have the ability to continuously learn and optimize based on user feedback and iterative training with further logs, KPIs and traces, promising even greater gains.”
