Service
Private AI deployment for Canadian businesses
We run language models on hardware or private servers you control, so client files, contracts and prompts stay inside systems you manage.
Who it's for
Businesses that can't send their files to a public AI service
- Accounting and bookkeeping firms with client financial records.
- Law and notary offices with privileged files.
- Clinics and health practices with patient information.
- Engineering and development firms with contracts, drawings and bids.
Why we take this seriously
Data control comes first
Our founder built Obsideo, an encrypted storage network where customers hold their own keys and the operators storing the data can't read it.
We bring the same approach to private AI. We settle who controls the data first, then build the system around that decision.
Where it can run
Three ways to deploy, compared
| Option | Where your data goes | Suits | Trade-off |
|---|---|---|---|
| A server in your office | Stays on your own network. | Small teams working with sensitive documents. | You buy the hardware and need a plan to maintain it. |
| A private cloud server | Stays in an environment you rent and control, in a Canadian region if you require it. | Staff in more than one location. | GPU servers carry an ongoing monthly cost. |
| A hosted AI service | Goes to the provider under its business terms. | Work where model quality matters more than isolation. | You depend on the provider's terms, pricing and availability. |
We include the hosted option because it's sometimes the better answer. Many projects use a mix of the three.
What a deployment includes
The parts around the model
Model testing
We compare models on samples of your real work, then choose one based on those results.
Document search
Staff ask questions in plain language and get answers with links to the source documents.
Access and logging
Sign-in, permissions and usage logs are set up to match who should see what.
Upkeep plan
We document backups, updates and how to change models, so the system stays maintainable.
Common questions
Are local models as good as the big hosted ones?
Not on every task. The strongest hosted models still do better on some work, while open models handle many routine jobs well. Testing on your own documents is the only reliable way to tell, so we do that before you buy hardware.
What hardware do we need?
That depends on the model, the number of people using it and how fast it needs to respond. We size it after testing and give you the options before you spend money.
Does private hosting make us compliant?
Not on its own. Controlling where data goes can help you meet obligations under laws such as PIPEDA and BC's PIPA, but it doesn't settle them. Your legal adviser should make that call.
Can it work with our existing software?
Yes, in most cases. A private model can sit behind the same kinds of integrations as a hosted one.
Working with information that can't leave the building?
A short description is enough. We'll reply with whether we can help and what a sensible first step would be.