Certified Centurion · Implementation

Your own AI, built, installed and running in your building.

Centurion is a private AI platform that runs on hardware your company owns, on its own data, with a model choice that is not locked to one supplier. I designed and build it, and I run my own business on it.

As a certified Centurion implementation partner I do the physical and technical work end to end: I specify the machine, build it, install Centurion on it, configure it around how your business actually works, and hand it over. You own the hardware, the data and the decisions. There is nothing to rent back from me afterwards.

Adrian Barkus building a Centurion machine — fitting the power supply into an open case on a workbench.
Building the machine. The hardware is bought in the client's name and installed on their own network — it is theirs from the first boot.

When this is the right answer

This is a sovereignty purchase, not a cost saving.

Worth saying before anything else: if your work can go to a cloud AI provider, a subscription will almost always cost you less than a machine and an installation. Buying your own is not the cheap route and I will not pretend otherwise.

It is the answer for the work that cannot legally or contractually leave the building — legal privilege, client confidentiality, patient and clinical records, deal material under NDA, anything caught by data residency or GDPR obligations. For that work the alternative is not a cheaper subscription; it is not doing the work with AI at all.

If that is not your situation, tell me and I will say so — a cloud subscription is the right recommendation and I would rather give you that answer than sell you a server you do not need.

What the installation involves

Six steps, from an empty desk to a system your team runs.

01

Specify the machine — and not more of it than you need

Centurion runs on two boxes, and they do different jobs. The Brain is a small, always-on computer that runs the agents and the daily work. The Core is a mini server with real storage that runs the language model itself. Most companies start with the Brain and add a Core when the data genuinely cannot leave the building. I match the specification to the workload rather than the brochure. You buy the hardware in your own name, so it lands on your asset register, under your warranty, from your supplier.

02

Build and prepare it

Linux, hardened to behave as an appliance rather than a workstation, named and addressed on your network, sitting in your building. It is yours from the first boot — there is no account of mine it depends on, and no dashboard of mine it reports back to.

03

Install Centurion on it

The platform, the model layer and the phone Remote, installed and configured rather than left on defaults. Where there is a decision to make — which model, how much context, what runs locally — I make it with you and write down why.

04

Load your knowledge

Your documents, your precedents and your way of working become what it answers from. This is the step that makes it yours rather than a generic assistant with your logo on it, and it is the one most often skipped.

05

Set the rules

Who can ask what. What it may read. What it keeps, and for how long. What it must never see. Governance is configuration here, not a policy document nobody opens — the boundaries are enforced by the system rather than trusted to good behaviour.

06

Hand over

Your people are shown how to use it, the runbook is written down, and the machine keeps working when I am not in the room. That is the point of the exercise.

Configured around you

“Standard installation” is not a phrase that applies here.

The machine
Specified to the workload — a small appliance for the daily work, a server with storage and a GPU when the model itself has to run on site.
The model
Chosen for the job and kept swappable, so your capability is not rented from one supplier under terms that can change without notice.
Your data
Your documents and precedents are what it answers from, on your premises.
Your workflows
The jobs it does are the jobs your people actually have, not a demo of what an AI could theoretically do.
Access
Who may use it, what they may reach, and what is logged. Configured per role.
Where it lives
On your network, at your address, on your power. Data residency stops being a risk you manage and becomes a property of the architecture.

The platform

Nothing in this system reports back to me.

The machine is bought in your name and sits in your building. The data stays on it. The model choice stays yours. When the engagement ends, what you have is a working system your team operates — not a dependency on me, and not a subscription that stops the day you stop paying. Centurion is licensed per user and per year; the hardware and this installation are bought and delivered separately, which is what keeps the two honest.

See the Centurion platform

How an engagement is shaped

A defined deployment, or ongoing support across the estate.

Contract — a defined deployment

Scoped, specified, installed, documented and handed over. You finish the engagement owning a working system and the knowledge to run it.

Retainer — ongoing support and adoption

For companies spreading the platform across departments: further deployments, model changes, new workflows, and someone accountable when something needs diagnosing.

The right starting point is usually a conversation rather than a quote. Tell me what the data is and who needs it, and I will tell you whether this is the right machine — or whether you should not buy one at all.

Start with the work that cannot leave the building.

Bring me the job you have been told an AI cannot touch because of where the data sits, and we can find out whether that is still true.

← Back to home