Sovereign, on-prem AI for network operations
Your engineers, shipping working AI agents on your own network — in one day.
Data never leaves the building. No working agents, no payment.
Private workshops and 90-day autonomous-NOC pilots for telcos, MSPs
and regulated enterprises that can't send their data to the cloud.
Why now
The AIOps tools that could help you aren't allowed in your building.
Cloud AI wants your configs, your telemetry, your topology. In a regulated NOC, that's a non-starter. So the toil stays manual — and the alerts keep winning.
The alerts never stop
Night-shift triage, repeat incidents, tribal knowledge locked in a few senior heads. Your best engineers spend their hours on work an agent should own.
Cloud AI can't come in
SaaS AIOps means shipping sensitive network data off-site. For telcos, finserv, public sector and critical infrastructure, that door is closed by policy — and by regulators.
Generic "AI training" doesn't ship
Slideware and ChatGPT-for-business workshops leave you with notes, not running systems. Your team needs agents working on their stack, not a demo on someone else's.
Autonomy without guardrails is a liability
Nobody sane lets an LLM push config to production unchecked. Without pre-validation and human-in-the-loop, "autonomous" is a word that ends careers.
How you buy
Prove it on your stack. Then scale what works.
One path, three steps. You never commit to the big number before you've seen agents running on your own network.
Prove it in a day
€4,900 (1 day) · €7,500 (2 days)Up to 12 of your engineers build alert-triage, config-generation and alert-to-action agents on your own stack. You leave with working agents — guaranteed.
See the workshop →AutoNOC 90-day pilot
from €25,000One agent, one workflow, one success metric — agreed in writing before we start. Runs on your hardware. A clear yes/no in 90 days.
See the pilot →AutoNOC license
Annual · on requestOnce the metric is hit, roll AutoNOC across workflows and sites under an annual license. Fully on-prem, your data, your control.
Talk scope →The product · AutoNOC
Autonomous NOC operations that run entirely inside your walls.
Five specialized agents on fine-tuned open-weight models — no cloud calls, no data egress. Every proposed change is pre-validated against a digital twin before a human ever sees it.
Five-agent flow, plain language.
Digital-twin pre-validation
Every change is tested in a Containerlab mirror of your network before a human is asked to approve it. No blind pushes to production.
Deterministic execution catalogue
Agents don't improvise commands. They select from a vetted, versioned catalogue of known-good actions — auditable and repeatable.
Tiered human-in-the-loop
You set what runs automatically and what waits for sign-off, by risk tier. Autonomy you can dial up as trust is earned.
On-prem small language models
Fine-tuned open-weight SLMs on your hardware. Air-gap capable. Your data is never a training set for anyone else.
The fastest way in · Private team workshop
In one day, your team builds AI agents that run on your own network.
Not a lecture. Up to 12 of your engineers build three working agents on your stack and workflows:
Alert-triage agent
Cuts the night-shift noise down to what actually needs a human.
Config-generation agent
Turns intent into vendor-correct configuration your team reviews and ships.
Alert-to-action agent
Closes the loop from signal to validated, approved change.
What it costs. What you leave with.
Risk reversal
If your engineers don't leave with working agents, you don't pay.
We carry the risk, not you. The workshop is built on your own stack, so "it works" means it works where it counts — on your network, in your hands, before you spend a cent on a pilot.
Proof, not promises
Practitioners showed up. So did the stages.
Also: the free, config-aware CVE scanner at scanner.vexpertai.com and Gen AI for Network Engineers — Volume 1 out now, Volume 2 in August.
Start free
See the work before you ever book a call.
🛡️ Config-aware CVE scanner
Point it at your platform and version. It tells you which CVEs actually apply to your config — not a generic feed. Live, free, no login wall.
Run the scanner →🗺️ Cairn — full-stack posture
See vulnerabilities, misconfigurations and exposures across your entire stack — cloud, hosts, apps and network. Designed to run on-prem; try the hosted demo, free.
Open the Cairn demo →📘 Gen AI for Network Engineers
Volume 1 and its 12 hands-on Colab notebooks — from your first API call to production. Volume 2 lands in August.
Get the notebooks →Who you'll work with
Eduard Dulharu — founder & CTO.
20+ years architecting enterprise and provider networks and security — AT&T, Infosys, Kyndryl — CCIE-level, MSc IT Security, six years inside NATO environments. Lecturer at Politehnica Bucharest and author of Gen AI for Network Engineers (Volume 1 out now, Volume 2 in August). Pilots and workshops are led personally.
One call. Fifteen minutes.
Tell us your environment. We'll tell you if this fits.
No deck, no interrogation. You describe your network, your timeline, and what your board is asking for — we tell you whether a workshop, a pilot, or honestly someone else is the right call.
Book a 15-minute fit call →15 minutes · no obligation.