Unryo adds agentic AI to link RAN faults with Kubernetes outages

Unryo has launched an agentic AI layer that correlates radio access network faults with the Kubernetes infrastructure running network functions beneath them. The tool is meant to speed root-cause analysis during outages by assembling evidence from both domains in one explanation.

Why it matters: - Cloud-native radio access networks split fault evidence across telecom and cloud tools, which can slow outage response. - Unryo's approach aims to reduce bridge calls by giving operators one reproducible explanation that ties radio symptoms to infrastructure issues. - The platform is designed to keep telemetry and incident data inside the operator's environment when deployed on-premises or in a private cloud.

What happened: - Unryo released an agentic cross-domain investigation capability for RAN faults and the Kubernetes infrastructure running the network functions beneath them. - The company says the agentic AI layer performs the same steps an engineer would use to investigate an alarm. - The feature is available now and can be deployed on-premises or in an operator's private cloud.

The details: - In a cloud-native RAN, network functions serving a cell run as pods on a container platform. - When a cell degrades, the root cause can sit in either the radio domain or the Kubernetes layer. - The agent reads the alarm, walks the dependency path from affected radio objects to the network functions, then to the pods and nodes running those functions. - The system retrieves state at the moment of the fault. - It checks radio counters and service-based interface signaling records from probes against platform events. - The output is a written explanation of what failed and why. - In one example, the agent traces a throughput drop to packet loss on the Kubernetes node beneath the cell, then treats the pod restart as a consequence. - Unryo's Topology Data Fabric powers the reasoning with a live model built from operator sources, including Nokia MantaRay, Ericsson ENM, the Kubernetes API, and SNMP or gNMI data from transport devices. - The model updates as the network changes. - The correlation is deterministic, which makes conclusions reproducible after an incident. - The agentic AI layer is model-agnostic. - Unryo's social media link is More information.

Between the lines: - The release positions AI less as a generative assistant and more as a guided investigation layer tied to live operational data. - That framing matters because operators need answers they can defend after outages, not just plausible summaries. - By stitching together telecom and infrastructure evidence automatically, Unryo is targeting the gap where cross-domain blame often delays fixes.

What's next: - Operators can deploy the platform without moving telemetry outside their own environment. - The main test will be whether the agent can consistently shorten investigation time across different vendors, network setups, and fault types.

The bottom line: - Unryo is betting that deterministic, cross-domain AI can replace manual reconciliation between RAN and Kubernetes teams during outages.