About
An independent AI practice that works inside other people's perimeters.
QAI Labs is the applied AI practice of Sabia Solutions Ltd, working from Glastonbury across the United Kingdom. We have been building and running engineering teams' infrastructure since 2021.
Most organisations that cannot send their data to a third-party API are told to wait. The architectures on offer assume egress, the pricing assumes tokens leaving the building, and the reference implementations assume a managed service in someone else’s region.
We exist to build the alternative. Sovereign AI, in the sense that matters operationally: the system runs on infrastructure you control, the data and the state stay inside your boundary, and the model is a component you can replace rather than a dependency you are married to. That includes on-premises, private and sovereign cloud, and genuinely air-gapped enclaves.
Sovereignty is not the same as isolation, and treating it that way produces worse systems. The useful question is which parts of the stack have to stay under your control and which do not. Weights, state, logs and evaluation evidence usually do. A general-purpose model behind an API sometimes does not. We would rather work that out with you than sell you the maximal version of it.
We are not a reseller, we hold no vendor partnerships that would bias a recommendation, and we do not sell hardware. Where an existing product would serve a client better than a build, that is what we say, including when it costs us the engagement.
Where this came from
The AI practice is new. The engineering underneath it is not.
- 2021
TecPeople Services Ltd
The practice starts as a technology consultancy, delivering platform, infrastructure and DevOps engineering into enterprise environments where the constraint was always somebody else’s change process.
- 2024
Sabia Solutions Ltd
Incorporated for business administration and business management consulting, and now the company that carries the group forward. TecPeople is winding down in an orderly way, with obligations settled first.
- 2026
QAI Labs
The applied AI practice, focused on sovereign AI: systems that run on infrastructure the client controls, with the governance and evaluation evidence that regulated work demands.
How we decide things
Evidence over assertion
Every claim we make about a system should be traceable to a measurement somebody could have run and got a different answer from. Where we do not have that, we say so.
Handover, not lock-in
The exit plan is written during Frame and tested during handover. Follow-on work should be a choice, not a consequence of dependency.
Constraints first
We would rather start from what you cannot do. The constraint determines the architecture; the use case rarely does.
Publish the limitations
What a piece of work does not show is the section we would want to read first in someone else’s, so it is the section we write in our own.
Say no clearly
Declining work we are wrong for is cheaper for everyone than discovering it in month three. We keep a written list of what we do not do.
Run it ourselves first
Our own operations run on a sovereign agent we built and host. Anything we recommend to a client has usually cost us something to learn.
Start with the constraint.
Most of these projects are shaped by what you cannot do rather than what you want. Data that cannot leave the estate, a model you cannot host with a third party, a decision somebody has to justify to a regulator. Tell us yours and we will say honestly whether we can work inside it.