AI that ships
Wired in, not bolted on
MCP servers, LLM features, and retrieval — inside the system you already run.
Where AI speeds things up — without cutting corners on quality.
Your tools and data, reachable by an AI agent. Built against the real workflow.
Search, summarising, classification — against your data, with failure modes handled.
Answers grounded in your own documents, with the source attached.
The people who own the outcome stay in the loop. Every time.
AI earns its place when it sits inside the system your team already uses — and when someone stays accountable for what it produces.
Most useful AI is not a chatbot bolted to the front page. It sits inside the system your team already uses, doing a job someone was doing by hand.
MCP servers
Your own tools, your own data, wired into an AI agent. Not a copy-paste workflow.
LLM features in the product
Search, summarising, classification, drafting — built against your data, with the failure modes thought through.
Retrieval over your own content
Documents, tickets, catalogues. Answers that cite the source instead of inventing one.
Agent tooling for the team
Automation aimed at the repetitive part of the work, reviewed by the people who own the outcome.
An MCP server that keeps team documentation in Atlassian and Confluence up to date — AI-agent tooling wired directly into an existing workflow, not a bolt-on chatbot.
Winner of Team Alpha at Sprint Zero, Techleap's first AI Hub event — a single day, real challenges brought by large Dutch companies, prototyped with Claude and MCP. A jury member said the winning solution solved a clear need well enough they would consider buying it in production.
I don't train models or build ML pipelines from scratch. The certifications are there — Stanford's Machine Learning, the Deep Learning Specialization — but that is not what I ship.
What I ship is production software with AI wired into it, by someone who has spent 20 years keeping production software alive.
The tools I use daily.