Praetava is developing dedicated GPU capacity, private inference and an onshore data platform for government and regulated workloads. We're not open for service yet. This is what we're building, who it's for, and the standards we're building it to.
Our mission
No Australian public-sector team should have to choose between the compute their work needs and the control their obligations demand.
AI capability has mostly been built somewhere else, on infrastructure answerable to someone else. For teams handling sensitive information, that has meant scaling back the ambition or carrying risk that never shows up on a balance sheet. Our purpose is to remove that trade-off — by operating accelerated compute here, to the standards the Commonwealth already sets, so that “where does our data go?” has a short and verifiable answer.
What we're building
The services below describe the platform as it's being designed for production. Each one is marked with where it currently stands — an early working prototype, or still on the drawing board.
Dedicated, single-tenant accelerated nodes with no hypervisor between the workload and the silicon. Root access, a fixed hardware manifest, and a boundary you can draw on a page.
Private, metered model endpoints for teams that need capability rather than hardware. Token-based billing, no shared inference pool, and no prompt or output data leaving the country.
Bring an open-weight or in-house model and we run it inside your boundary — deployment, serving, scaling and versioning, without handing the weights to a third party.
A lakehouse and model-serving control plane for teams that can't send data offshore. Onshore storage, governed access and lineage, feeding the same fleet the models run on.
Platform
We've deliberately drawn the line above the application layer. No agents, no copilots, no chat products — that's your work, and your users' work. What we intend to own is everything underneath it: the serving layer, the data platform, the fleet, the network and the facility those things sit in.
Owning that depth is what will let us answer a question most providers have to escalate: for any given workload, which systems hold the data, under which certification, and within which controls.
Who it's for
If your architecture review starts with residency and ends with an authorising officer's signature, we're being built for you.
Programs handling sensitive or PROTECTED information that need accelerated compute kept onshore.
Delivery partners who need a compliant compute layer beneath a solution they're accountable for.
University and publicly funded groups working on data that can't be moved to an offshore hyperscaler region.
Health, financial services and critical infrastructure operators facing equivalent residency and audit obligations.
Standards
Retro-fitting compliance onto infrastructure is how early providers end up unable to sell to government. So we started from the control set rather than the hardware, and we'd rather state our position plainly than display badges we haven't yet earned.
The government workloads we’re building to support require hosting in facilities certified under the Hosting Certification Framework, in line with applicable Australian Government hosting requirements
Why provider selection comes first
Origin
Praetava started from a recurring observation: accelerated AI workloads kept stalling before they reached production, and rarely because of the model or the budget. The blocker was that the compute on offer couldn't be placed where the data was required to stay.
We're a small Canberra team, building the company privately while we design the platform and select the certified facility it will run in. Our backgrounds are in infrastructure engineering and delivering technology into regulated environments.
We think about infrastructure the way you'd want a foundation built — unglamorous, structural, and the reason everything above it stays standing. That's the whole job, and it's the standard we hold the platform to.
Contact
We're shaping the first release around real workloads. If you have one that can't currently be placed — the classification, the rough size, the deadline — we'd like to hear about it, and we'll tell you honestly whether and when we could carry it.
hello@praetava.com