New Aether Paper on Agentic AI Network Validation

Our paper Aether: Network Validation Using Agentic AI and Digital Twin is now available on arXiv as arXiv:2604.18233.

Aether is a framework that lets network operators express high-level validation intents in natural language. A multi-agent system autonomously decomposes those intents into executable verification plans and runs them against a digital twin of the network — checking configurations, simulating topology changes, and verifying compliance with operational policies.

We demonstrate the approach on realistic scenarios derived from production ISP deployments, showing that it can identify configuration errors, predict the impact of planned changes, and accelerate troubleshooting workflows.

Joint work with Sam Betts, Giovanna Carofiglio, Giulio Grassi, Martin Gysi, and John Kenneth d’Souza at Cisco Outshift.