We build AI that keeps working when the network does not
RAPTOR AI is a UK company that builds RAPTOR: ruggedised, air-gapped edge AI for organisations that cannot send their data to someone else’s infrastructure. The product is the unit and the software that runs on it. The principles below shape how both are built.
How the product is built
Four principles that shape the unit and the software on it.
- Offline by design
- The unit is built to run with no network at all, from first boot onwards. If a feature would need a connection, it does not ship.
- Open weights, held by you
- The models are open-weight files on your device. You can archive them, inspect them and run them for as long as the hardware lasts. No inference API sits in the path.
- No telemetry
- There are no analytics, no online licence check and no update ping. What happens on the unit stays on the unit.
- Built for the operator
- The people using RAPTOR in the field are usually not platform engineers, so deploying a capability, testing it and checking its health all happen on one screen, without a terminal.
How an engagement runs
- 01
Workload first
We start with what you need to process: how many people, what kind of material and how long the documents are. That decides the configuration and the quotation.
- 02
Evaluation on real material
Wherever your security posture allows it, the capabilities are evaluated against your own material rather than a public leaderboard. A model that tops a public benchmark can still do badly on your documents.
- 03
Delivery, imaged and sealed
The unit arrives with the platform, the catalogue and the model weights already installed. There is no first-run download, and no account to create.
- 04
Updates and support on your terms
Under the annual agreement, new models and platform versions are delivered offline and applied on your schedule through your own transfer process, with support for your operators throughout.
One unit, one agreement
RAPTOR is supplied as the hardware unit, together with an annual licensing, maintenance and support agreement. Every configuration is quoted for your workload.
The RAPTOR unit
Case, console and two compute nodes (three on request), delivered with the platform and every model installed.
- Configured to the people and material you need to process
- Operator handover on delivery
Licensing, maintenance and support
An annual agreement covering the software licence, updates delivered offline, and support for your operators.
- Renewed annually
- The unit stays on your site throughout
There is no per-request or per-token billing.
Request a quoteTell us what you need to process
We size and quote a configuration to your workload and, where your security posture allows, set up an evaluation on your own material.