01
Private by architecture
Data routes, model access and retention are designed before deployment. Nothing leaves the agreed boundary because the system has no route for it to leave by.
Build · Private AI Systems
The model stays where the data belongs.
Models, inference hardware and controls deployed around the institution’s boundary, not a vendor’s appetite for data.
Overview
We build and operate AI systems for organisations that cannot send sensitive work to an unknown model behind a public endpoint. The model, hardware, data path and access boundary are specified together. Deployment can sit on dedicated hardware in the client’s chosen region or environment, with every input, output and change governed by an agreed control.
What it includes
The work, named plainly.
02
Hardware with an owner
Dedicated inference hardware sized, deployed and monitored for the workload. Location, access and responsibility are named rather than inherited from a cloud default.
03
Models chosen on evidence
Open-weight and purpose-trained models evaluated against the real task. We measure accuracy, refusal and failure modes before choosing scale.
04
A person remains accountable
Outputs that affect records, decisions or people pass through explicit review. The audit trail shows what the model produced and who accepted it.
05
Proof small enough to inspect
We designed and trained our own 12.2 million parameter language model. It runs entirely on the reader’s device.
Run the modelHow we work
Engineered once. Maintained indefinitely.
- 01 Classify the data and define where it is allowed to move
- 02 Measure the task before selecting a model
- 03 Design the hardware, network and access boundary together
- 04 Evaluate accuracy, refusal and failure against written cases
- 05 Deploy with logging, human review and a controlled rollback
- 06 Operate the system and re-approve material model changes
Speak with us
Tell us where the data must stay.
We reply within one working day, from an engineer who understands the deployment.
Send securely