What we do
Tailor-made platforms,
built to hold a record
One team, from the workstation to the server. We start from the real process — the one that lives on paper or in a spreadsheet — and turn it into a system that does the work and keeps the proof of what it did.
The company
IMGIAINFO inc.
A Québec company. Its name is made of three words — image, AI, information — which are also its trade: read an image, understand it with a model, and turn it into information that holds.
We work where data has to stay true: manufacturing processes, controlled records, registrations that must stand in front of an inspector. What we deliver is installed at the client's site, on their hardware, under their rules. The name states the method, and the method explains the name.
The fields
Four skills, one chain
1. Applications and web
Desktop and web applications, databases, data-entry interfaces, replacing paper with entry that leaves a trace. The workstation, the server, the network: we handle all three.
2. Data integrity and audit trail
Our main field. Every write carries who acted, from where, what was done, the value before, the value after, the reason, the timestamp and the result — and nothing can be erased. See the ALCOA++ page.
3. Trusted AI
Vision models (reading documents, forms, scans) and language models installed inside the work chain. Always by the same rule: the model proposes, the human validates, the log keeps both. See an LLM in the loop.
4. Sovereign cloud
On your hardware, inside your network, or with a provider you choose — public, private or local, according to your rules. Data does not leave to a third party that never asked your permission. See digital sovereignty.
How we work
Four principles, held in the code
The software must do better than paper
Not the equivalent: better. What paper did badly — saying who wrote what, preventing a rewrite without a trace — the system must do without room for argument.
Rules are data, not code
What varies from one client to the next, or one product to the next, lives in tables that can be changed and traced — not in a hidden line of program.
One function per thing
Computing, drawing and checking the same object go through the same code. Two copies of the same formula are two ways to drift apart — and a measurement that no longer means anything.
We measure the result, not the intention
What is delivered is read and measured as it comes out, never as we believe we wrote it. A verified file beats a file assumed to be good.
Scope
What we do not do
- No automatic decision on a regulated record. No value written by a model without a named person validating it.
- No imposed dependency. No public cloud by default, no third-party service whose output you do not control.
- No client data in a showcase. What we build for a client lives at that client's site, under their rules.