Two engineers, a small number of deployments at a time. Currently taking on new work.
Services
You buy a machine, or we rent you a dedicated one. We install the model, wire it to your documents, and manage it over an SSH channel you control. Your team gets a search bar and a chat box. Nothing leaves your network.
What we actually do
You buy a machine, or we rent you a dedicated one. We install the model, wire it to your documents, and manage it over an SSH channel you control. Your team gets a search bar and a chat box. Nothing leaves your network.
FIG.1
01
One call, then a spec instead of a slide deck.
We look at the work your team actually does and which parts are blocked today. You get back a shortlist of use cases worth building, the model that fits them, and the exact hardware it needs, priced.
FIG.2
02
Your hardware, or a dedicated GPU we rent for you.
Buy a box and keep it in your own rack, or let us put a single-tenant GPU in an EU data centre in your name. Either way we install the model, the document search, and the internal API, and we manage it remotely over SSH.
FIG.3
03
No third-party API anywhere in the request path.
Inference happens on your machine. There is no call out to a model provider, so there is no processor agreement, no transfer, and no retention question. Every query is logged inside your own perimeter.
FIG.4
04
Monitoring, patches, and model upgrades on one monthly fee.
AI systems are not set and forget. We watch performance, apply updates, and swap in better open models as they ship. When something breaks, the alert goes to us and not to your service desk.
The decision on one page
This is the table your security officer will build anyway. Here it is up front.
| Criterion | Cloud AI | SRI |
|---|---|---|
| Where inference runs | Vendor infrastructure, region of their choosing | Hardware you own or rent in your own name |
| Who holds your documents | A third party, under their retention policy | You. The index never leaves your disk |
| Processor agreement needed | Yes, plus transfer assessment | No third-party processor in the request path |
| Works with no internet | No | Yes, on-premise and air-gapped |
| Audit log location | Vendor console, exportable at best | Your own log stack, queryable like any other service |
| Model choice | Whatever the vendor ships and deprecates | Open weights you pin, upgrade, or roll back |
| Cost shape | Per token, unbounded, scales with success | Fixed hardware or fixed monthly, flat under load |
| If the relationship ends | Access stops, data export on their terms | The system keeps running. You already own it |
What it actually gets you
Ask a question in plain language and get an answer with citations from your contracts, policies, and archives.
First-pass drafting, summarising, and clause comparison on documents that were never allowed near a cloud tool.
Classify and route tickets, mail, and forms automatically, with a human check where it matters.
Every prompt and answer logged inside your network, attributable to a user, retained on your terms.
What it costs
The reason cloud AI budgets explode is that the bill scales with adoption. This does not. You pay to have it built, you pay for the machine, and you pay a flat fee to keep it running.
Workflow review, model and hardware sizing, integration plan, and the written data flow your security team signs off on. Credited against the deployment if you go ahead.
Fixed fee, one off
Everything installed, wired to your documents and systems, access control and audit logging in place, runbooks handed to your IT team. Quoted before you commit.
Fixed fee, one off
Buy the hardware and own it outright, or rent the same shape from us as a single-tenant GPU in an EU data centre. We tell you honestly which is cheaper for your volume.
Your capex, or our monthly
Monitoring, patching, model upgrades, and capacity planning. Flat under load, because the cost of an extra thousand queries on hardware you already own is zero.
Flat monthly
Both fixed fees are quoted after the discovery call and before you commit to anything. There is no per-seat licence and nothing that expires.
Next step
A 45-minute call. We look at your workflows, your data rules, and whether local AI is worth it for you. If it is not, we will tell you that instead of selling you a project.
Not ready for a call? Send the security brief to whoever has to approve it, or just reply to an email. All three work.