Sovereign on-premise AI
Using an online AI service on patient records or contracts means sending them to servers you do not control. We install the language model on your premises, on your hardware. It reads your documents, answers your teams, and nothing crosses the door of your building.
What you gain
- 0 data sent to a third-party service
- 0 token-based billing
- 100 % of the model under your control
Using an online AI service on sensitive data means accepting three things: that your documents pass through a third party, that the bill depends on your usage, and that the terms can change without you. Local deployment removes all three.
We install the model, index your documents in a vector database that stays on your network, and impose access partitioning at engine level. The assistant answers in your language, cites its sources, and knows only what its user is entitled to know.
What the software includes
Local language model
<p>An open model served by Ollama on your own hardware, with no outbound network call.</p>
Document retrieval
<p>Your documents are indexed in a local vector database and become queryable in natural language.</p>
Partitioning by role
<p>The data boundary is applied before reading, not merely at display time.</p>
Voice transcription
<p>Dictated notes transcribed locally, with nothing sent to an external recognition service.</p>
Offline operation
<p>An internet outage interrupts neither the assistant nor document retrieval.</p>
Hardware sizing
<p>We assess the machine required before any commitment, based on the number of users.</p>
The embedded AI
DentalNova
The dental practice's administrative assistant, partitioned by role.
MedicalNova
The clinic's management assistant, partitioned department by department.
AssistNova
The website chatbot that answers from your real data, in real time.
PharmaNova
The batch file reconstructed in one question, the procedures made queryable.
ImagerieNova
The dictated radiology report, transcribed and filed on site.
ComptaNova
Documents filed, deadlines met, the client file found.
JuriNova
The case file, the procedural deadlines and the clause found in one question.
MatchNova
A reasoned matching score between an application and a job description.
DocuNova
Your internal documents, queryable in plain language, without leaving your walls.
Who uses it
Further reading
-
Law 18-07: where must health data be hosted in Algeria?
Law 18-07 governs the processing of personal data in Algeria and reserves a particular regime for health data. What that means in practice when choosing clinic software.
-
Running a language model locally: what to measure before buying the machine
Before choosing between a 32-billion and a 72-billion parameter model, you need to know what to measure. Here is the protocol we apply and the orders of magnitude to check on your own hardware.
-
Cloud or on-premise ERP for an Algerian clinic: the comparison
The choice between an online ERP and one installed in the establishment commits you on confidentiality, service continuity and cost over several years. A point-by-point comparison.
-
What machine do you need to run an AI locally?
Video memory decides almost everything. Here is how to size a machine to run a language model on your premises, and how to check for yourself before buying.
Frequently asked questions
Do you need a powerful server to run AI locally?
Less than people think. A workstation with a recent graphics card, or a Mac with an Apple Silicon processor, is enough to serve a language model for a team of ten to thirty people. We size the hardware during scoping, before any commitment.
What happens if the internet goes down?
The software keeps working. Since everything is installed on your local network, losing the external connection interrupts neither data entry, nor the AI, nor document printing.
Can you migrate our existing data?
Yes. We take on spreadsheets, exports from existing software and historical databases. Migration is a dedicated stage with a consistency check before go-live.
Do you build the mobile app as well?
Yes, in Flutter, with a single code base for Android and iOS. The app connects to the same server as the desktop software and is therefore subject to the same data boundary.
How do you guarantee a user only sees what concerns them?
Partitioning is applied in the data access engine, not merely in the interface. In practice, a user cannot obtain information outside their scope by rephrasing the question to the AI assistant: the restriction is imposed before the data is read.
Updated Aug. 11, 2026