About
My days are spent keeping the IT of a rubber manufacturing plant alive —
servers and networks, identity and access for everyone who works there, and the
internal web applications that production, purchasing, quality and HR rely on.
There are a little over forty systems in that inventory, from the ERP to the
camera software, and most of them now sign in through one door.
Over the last year the work moved closer to the machines. Signals from PLCs on
the mixing and pressing lines now arrive as structured data: who is on shift,
which work order is running, why a line stopped, what an hour of production
actually cost. It is unglamorous plumbing, and it is the part I enjoy most.
AI runs through all of it. Some is off the shelf — OpenAI and Anthropic
models, retrieval over our own documents, n8n workflows that stitch systems together. Some is ours, hosted on our own hardware
where a public API is too slow, too expensive or not allowed to see the data.
And some of it now does real engineering work: agents read a request, study the
repository, propose a plan, and open a pull request. A person still decides what
gets merged.