Knowing which model fits which job
Claude, GPT, Gemini and open-weight models, each chosen on strengths, limits, cost and context. Per task, not per hype cycle.
I'm Steven de Graaf, project engineer for hydraulic, oxygen, ozone, water and compressed-air installations, from first concept to commissioning. AI builder by choice.
PEFA deNOx is a new deNOx installation at Tata Steel's pellet plant in IJmuiden, built on technology that has not yet been applied at this scale anywhere else in the steel industry. It's a large multidisciplinary project. I'm one of the project engineers, responsible for the utilities package: all the media that feed the new installation.
Concept
Cost estimate
Engineering
Construction
Commissioning
* Combined investment in the dedusting and deNOx installations at the pellet plant, as published by Tata Steel Nederland.
A pipe network moves gases and liquids. A neural network moves information. Both come down to flow, pressure, feedback and failure modes, so the instincts that keep a plant running translate directly to AI.
Claude, GPT, Gemini and open-weight models, each chosen on strengths, limits, cost and context. Per task, not per hype cycle.
I use Claude Code and Codex as a daily engineering team: I direct and review, the agents do the typing.
I use image, video and interface generation to prototype ideas in hours instead of weeks.
Real products, not demos. Next up: mijnoefenplan.nl.
This site was designed and built with AI agents, directed by me.
I enjoy the moment an idea on paper becomes something that actually works in the field. I take ownership, make the call, and carry a project from first sketch to final commissioning.
Off the clock, you'll find me on the climbing wall or on my bike.
Press and hold the handwheel. Or don't.