About
A software engineering company for systems that include AI.
Cognitum Technologies is a French engineering company that designs, builds and evaluates software and AI systems for industrial companies, software editors and organizations with sensitive data. A senior team with more than twenty years of professional practice, and a product line launching in 2027.
- +More than twenty years of professional software engineering
- +Six capabilities: architecture, LLM, agents, AI systems, evaluation, MCP
- +Own GPU inference infrastructure in France
- +A trained personal-data detection engine
01 — What we do
We build AI into software, and we build software with AI.
An AI feature is only as good as the software it lives in and the evaluation that proves it. So we cover the whole chain, with the same engineers.
Architecture and development process consulting
Reviews and target architectures for cloud, web and mobile platforms, and the delivery process around them: pipelines, testing, review, AI-assisted development.
LLM engineering
Retrieval (RAG), embeddings, structured extraction, context engineering, model selection across hosted, open and private models.
Agent engineering
Agents built with the OpenAI Agents SDK and the Microsoft Agent Framework, with tools, approvals and tracing.
Building AI systems
Pipelines, anonymization, orchestration, hosting and run, in the public cloud or in private environments when the data requires it.
AI evaluation
Evaluation sets, metrics agreed with the business, regression gates and monitoring in production.
Tool integration with MCP
MCP servers over business systems, with scoped permissions, so assistants and agents can act safely.
These are the skills of what is now called AI-forward engineering: engineers who use AI to build, and who build the systems AI runs in. We practise both every day, and we only ship what we can evaluate.
02 — Values
Four commitments on every project.
Simplicity
Sober architectures, easy to maintain by your team, with no component that is not earning its place.
Technical mastery
We understand in depth what we deliver, down to the models, rather than assembling tools we cannot explain.
Compliance by design
GDPR and AI Act obligations are built into the product and the process, not added afterwards.
Measurable results
Every project follows a business metric and an evaluation set: response time, volume processed, quality, conversion.
03 — Team
Senior engineers who have shipped production systems for more than twenty years.
Cognitum is led by a senior software architect with more than twenty years of proven professional experience in .NET and Azure, and thirty years of writing software. Every consultant who joins is a senior engineer in software architecture, cloud or applied AI, and specialist freelancers cover specific ERPs and front-end work. The team you meet in the workshop is the team that builds and runs your system.
Senior only
No junior bench. Each engagement is staffed with people who have shipped production systems.
One accountable architect
Every project has a named senior architect accountable for the design, whoever delivers it.
We are hiring
Senior .NET and AI engineers who want to build products for industry, in France. Write to us through the contact page.
04 — Assets already built
Foundations that make each project cheaper than the last.
Three pieces of technology we own and reuse across clients.
Models
Personal-data detection engine
Transformer models fine-tuned on commercially licensed data, exported for native .NET integration, with a fine-tuned open model for the hard cases.
Platform
Multi-tenant SaaS foundation
A modular .NET platform with authentication, fine-grained authorization, tenant isolation and dedicated isolation for regulated clients: the common base of our products.
Infrastructure
GPU inference server
Workstation-class GPU infrastructure in France for demonstrations, pilots and local inference, so no client data leaves the perimeter during a pilot.
05 — Where we fit
Organizations facing one of these situations.
Software editors and product teams
AI features to ship
A product that needs LLM features, agents or MCP integrations, built by engineers who deliver production code and the evaluation that goes with it.
Industrial SMEs and mid-caps
Processes to automate
Quote requests, leads and documents handled by hand, an ERP that is the system of record, and a need to integrate rather than replace.
Regulated and sensitive sectors
AI that must stay under control
Defense, health, energy and public bodies that need models hosted in France or on-site, and anonymization before anything else.
06 — Where we are
A product line funded by engineering work, launching in 2027.
Engineering and consulting missions for software, industrial and regulated clients; development of the commercial automation platform; discovery workshops with industrial companies.
Version 1 of the commercial automation suite: quote request processing, lead qualification, sales assistant.
Commercial launch and first early-access pilots with industrial companies.
Growth of the senior engineering team, ERP integrator partnerships, and new use cases on the same platform.
If your system needs AI that holds up in production, start here.
The best conversations take the real product, the real process and the real constraints seriously from the first workshop.