Run AI where your data has to stay
Three options, one rule: your data goes only where you have agreed it can go.
Your servers
Models, retrieval and agents run on hardware you own, with no outside connection if you need none. Best where policy forbids any data leaving the building.
- Full control of data and models
- Updates moved in through a process your security team approves
- You provide or buy the GPUs; we size them from load tests
Your cloud account
Everything runs in a cloud account only you control, in the region you choose. Faster to start than on-premise, with the same ownership of data and code.
- Fast to stand up and scale
- Data stays in your account and region
- Your cloud bill, with budgets and alerts we set up
Managed by Matrida
We operate the system, hold the review queue and send a weekly report. Reports, prompts and logs stay visible to you, and the system can move into your accounts on request.
- No tooling for you to own
- Monitoring, fixes and model upgrades included
- A clear exit: move it to your accounts later
How the three compare.
| Your servers | Your cloud | Managed | |
|---|---|---|---|
| Who holds the data | You | You | You, under agreed terms |
| Speed to start | Slowest, hardware first | Fast | Fastest |
| Who runs it day to day | Your team, with our training | Your team or ours | Matrida |
| Works with no internet | Yes | No | No |
Common questions
Which deployment option is right for us?
Start from your rules. If policy says data cannot leave your premises, on-premise. If it can leave but only into an account you control, private cloud. If you want the outcome without owning the tooling, managed. The strategy call settles it in about thirty minutes.
Can we change option later?
Yes. Systems are built from containers and infrastructure code, so moving from managed to your own cloud, or from cloud to on-premise, is a planned project, not a rewrite.
Which models can run privately?
Open-weight models of many sizes, and in some cases models from commercial providers under private-hosting terms. We choose by evaluation on your tasks, then by your rules.
What do you need from our IT team?
Access to the environment, a named contact for security review, and decisions on network and identity. We bring the architecture, the documentation and the tooling.
How do you keep a private system up to date?
Model, prompt and dependency changes go through evaluation first and roll out in stages. For air-gapped systems, updates are packaged and delivered through the process you approve.
Tell us where your data has to stay.
We reply with the deployment option that fits your rules, what it needs from your side and what it costs to run.
- A recommended option, with the reasons
- What your security team will ask, answered up front
- A rough size of the hardware or cloud bill