Local models on Edge
AI models running on-site on the more capable, server-class Edge hardware you control — not the small portable unit, but the same platform family we've shipped since 2017.
For work where the data can't leave the country, the building, or the device. We set up AI on infrastructure you control — local models on a Hyracbox Edge Server, your own hardware, or an in-country cloud — so nothing has to be sent to someone else's servers. This is the path for organizations ready to invest in AI that works fully offline.
Most AI can run in the cloud. Some can't. This is for the work where sending data out isn't an option — by law, by policy, or by conscience.
Patient records, child-protection cases, refugee identities or donor details that must stay in-country.
A regulator, a ministry, or a data-protection law that forbids sending the data to outside servers.
Sites with no reliable internet, where the AI has to keep working with nothing but local power and hardware.
A funder or board that needs to see, concretely, that beneficiary data never leaves your control.
We choose the option that fits your law, your sites and your budget — from a small on-site server to your own data centre.
AI models running on-site on the more capable, server-class Edge hardware you control — not the small portable unit, but the same platform family we've shipped since 2017.
On the servers you already run, in your building or your data centre, under your control.
Where a local cloud region satisfies the rules, we deploy there — data stays inside the border.
Data flows designed so nothing crosses the boundary. What's in stays in; nothing phones home.
Documentation a board, a donor or a regulator can read to see exactly where the data lives and moves.
Your staff trained and equipped to operate it — with support arranged for the parts that need us.
Running AI on your own infrastructure is a real investment — in hardware, setup, and the care to do it safely. We scope each one to your sites, your data, and the rules you answer to.
How much AI you're running. Bigger models, or several of them, need more powerful hardware on-site.
How many sites. Each location comes with its own hardware and setup.
How strict the rules are. Tighter security, or a formal standard to certify against, means more work.
The state of your data. Data that needs cleaning or organising first adds time before anything can run.
Resilience. Backup hardware and failover, for the places where downtime isn't an option.
Ongoing support. Keeping it running and up to date over time, if you'd like us on hand.
Most deployments start lean and grow only where the mission genuinely needs it. We'll always scope to what's actually required — and say so if something isn't worth the cost.
Honestly, not always — local models trade some raw capability for the guarantee that data never leaves. For many real tasks that trade is well worth it. We'll tell you plainly where it fits and where it doesn't.
Most organizations are fine on the cloud, or on a cloud region inside their country. Sovereign deployment is for the specific cases where the data genuinely can't leave. If you don't need it, we'll say so.
It's our offline-capable server platform — and it comes in a range of forms, from a small, portable unit up to full server-class machines. The compact version is the popular one, and it's ideal for delivering content offline. Running local AI models on-site, as this service does, uses the more capable server-class hardware — which is part of what the price reflects.
Often, yes — this kind of work usually fits a technology or infrastructure line in a proposal. We can help shape that before it's funded; see Grant Implementation.
Tell us what the rules require, and we'll map the honest options — including whether you need this at all.