Internal AI tools
On my fixed-term contract at Schneider Electric, I built three internal AI tools and led best-practice sessions. The MCP server and the RAG agent were neither put into production nor adopted.
What I built
- A custom MCP server for EAE, an industrial automation software. MCP (Model Context Protocol) is the protocol that connects an AI assistant to tools and data.
- An LLM agent that retrieves the relevant context from Confluence documentation (RAG).
- An AI code-review pipeline, with prepared instructions (preprompting).
Sharing it
- “Copilot Champion”: monthly best-practice sessions on AI-assisted development.