Skip to content
RAPTOR — Rapid Analysis Platform and Tactical Operations Rig
The platform

Everything the unit needs to run is inside the case

RAPTOR puts every model and the interface on hardware you own. Documents, imagery and audio are processed on the unit, and the results stay there.

Exploded schematic illustration of the RAPTOR unit: transit case, console, compute modules and battery pack
Schematic illustration of the unit and its modules.
The console

One screen for the operator

The RAPTOR console is the control plane on the unit. It lists every installed capability, deploys the one the operator needs in a single click, and shows its health, logs and configuration on the same screen.

One capability runs on a node at a time, so memory and power stay predictable and it is always obvious what is running. The standard two-node unit runs two at once, and a third node adds a third.

Capabilities

Installed on every unit

  1. 01

    LLMs

    Drafting, summarising and extracting from text, on large language models that run on the unit.

  2. 02

    Translation

    Text translation for operational documents, entirely on the unit.

  3. 03

    Transcription

    Speech to text across many languages.

  4. 04

    Image and document analysis

    Ask questions about a frame, a scene or a change between two, and turn scans and PDFs into searchable, structured text.

  5. 05

    Image generation

    Briefing visuals and synthetic training data, generated on the node.

+

Bespoke capabilities

Beyond the installed set, capabilities can be developed for a specific mission, data type or workflow and delivered to the unit the same way.

Data in, results out

Works with the feeds and files you already have

Files the unit can see
Footage, recordings and scans are read from storage the unit mounts on your network, such as the share a recorder or scanner already writes to.
Systems you already run
Software on your site network can send material to the unit and receive results back, without installing anything on the source system.
Material carried in
Removable media is how data crosses an air gap in practice. Anything brought in is scanned before it reaches the analysis stack.

Continuous ingestion from a specific camera, sensor bus or message queue is integration work scoped with you per deployment.

Security

Built for the boundary

Physically isolated
No network path is needed at any point, from first boot onwards, so there is no route out for data and no route in for an attacker.
Scanned on entry
Media from outside the boundary is checked before any model reads it.
Encrypted at rest
Model weights and the material you process are stored on the unit’s own encrypted storage.
Updated offline
New models and platform versions are delivered offline and applied on your own schedule. Nothing is pushed to the unit.
FAQ

Questions, answered

Yes. The models and the console are installed on the unit, nothing is fetched at runtime and there is no telemetry. A unit can be imaged, sealed and deployed into a facility that has never been connected.

LLMs, translation, transcription, image and document analysis (including OCR), and image generation. Bespoke capabilities can be developed for a specific mission and delivered the same way.

As the hardware unit together with an annual licensing, maintenance and support agreement. Pricing is by quotation, sized to the configuration and the workload. There is no per-request or per-token billing.

The unit keeps working. RAPTOR has no online licence check and no remote switch, because an operation should never depend on a licence server. The agreement is a contract, not a lock: if it lapses, updates and support stop. If you decide to stop using RAPTOR, we remove the software and model library, and the hardware stays with you.

New models and platform versions are delivered offline and applied on your own schedule, through whatever transfer process your security policy allows. Nothing is pushed to the unit automatically.

That depends on the workload: short requests support far more simultaneous users than long document analysis. Tell us how many people and what they are processing, and we will size the configuration to match.

Tell 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.

[email protected]Full specifications