AI engineering and agentic development
AI is part of how we build software and of what we build for clients. Our position is practical: AI is a tool that must earn its place with a measured result, and it must work inside the same rules for security, privacy and quality as any other software.
AI in our own engineering
AI coding agents in daily delivery, AI for code analysis, auditing, refactoring and testing, and assistants, RAG and MCP integrations built into business software.
Agentic development
AI coding agents such as Claude and GitHub Copilot plan a change, edit the code, run the checks and correct themselves, under the direction of a senior engineer who reviews every result.
Architecture and pattern analysis
We use AI to read large codebases, map dependencies and compare design options before a decision is made.
Automated code auditing
Systematic review of code for defects, security weaknesses and departures from the project's standards.
Legacy modernization
AI-assisted refactoring of older systems in small, verified steps, which keeps the risk of a migration low.
Test generation
Tests are written with AI support and reviewed by the engineer, so coverage grows together with the code.
Written rules for the agents
Each project carries written instructions for the AI agents: the architecture, the coding standards, what must never change without approval, and the checks to run before work is called finished.
AI solutions we build for clients
Every solution starts from one measured problem and follows the same rules: a person approves anything sensitive, privacy is designed in, there is no lock-in to one AI provider, and the monthly cost is controlled.
Assistants for customers and staff
Chat and email assistants that answer questions, take bookings and requests, and hand over to a person when needed.
Document and data automation
AI that reads incoming invoices, orders, forms and emails, extracts the data and passes it to your business systems.
Knowledge assistants (RAG)
Retrieval-augmented generation: an assistant that searches your own documents first, answers from them and shows its sources.
Integration with LLM APIs, MCP and agents
Language models connected to your databases, files and internal APIs through controlled, permission-limited interfaces, including the Model Context Protocol (MCP).
Model choice and hosting
Commercial models such as Claude, GPT and Gemini, or open models such as Llama, Mistral and Qwen, on a cloud of your choice or on your own servers, selected per task for quality, cost and data location.
Evaluation, training and AI policy
A test set built from your real cases and run before go-live, hands-on workshops for your team, and a short, clear AI usage policy.
Where the experience comes from
AI workflows at Vitec Auto
As lead full-stack engineer at Vitec Auto, on a large web-based software platform for the automotive industry, Fabrice introduced AI-augmented engineering: language models for architecture analysis, code auditing, refactoring of legacy systems and test generation, with strict quality and security standards kept in place.
EyeOne's own products
This website, our SaaS products and our browser games are developed with AI agents working under written project rules, strict type checking and human review.
Explore the other skill areas
Have a project where this experience would help?
Tell us what you are planning and we will tell you how we would approach it.
