From the Plant Floor to Industrial AI
It is somewhere past midnight. A reading has drifted out of range, the shift is staring at a trend none of them has seen behave quite like this, and the one person who would have known the cause in thirty seconds retired last spring. Nobody wrote down what he knew. It left with him.
I have watched a version of that night more times than I can count — in the lab, on the floor, in a maintenance bay. It is the reason I build what I build. And it is why my path into AI did not start with a computer-science degree. It started on the plant floor.
The non-traditional path
I have spent 12+ years in and around cement operations: quality and lab work, cement chemistry and XRF, maintenance and lubrication, SAP workflows, and safety and MSHA awareness. For a long time I just did the work. Then I started noticing the same thing everywhere — the knowledge that actually ran the place was not in any system. It was in people, shifts, binders, and memory, and it was hardest to reach at exactly the moment it mattered most.
I am not a senior software engineer, and I do not claim to be. I am an AI product builder and technical architect who ships through AI-assisted development: I design the system, set the guardrails, curate the domain knowledge, and drive the build to a working, tested result. Stating that plainly is the point — it is honest, and it is where the value comes from.
Why the overlap matters
Most AI builders have never set foot in a cement plant. Most cement professionals are not building AI. I sit in the overlap, and that overlap is the whole advantage. I know what 2 a.m. on a plant floor actually feels like — the pressure, the stakes, the cost of being wrong — and I know how modern AI can be shaped into something genuinely useful instead of generic AI fluff. You cannot fake the first half, and most people in AI do not have it.
What I build
That perspective drives CementOps AI: cement-specific decision support for safety, maintenance, MSHA compliance, operations, SAP workflows, and plant knowledge. Along the way I have built 100+ custom GPTs across those domains. The framing never changes — it is decision support that strengthens experienced people's judgment, never a replacement for engineers, supervisors, legal counsel, or site procedures.
Why now
AI agents and custom GPTs finally make it possible to organize specialized knowledge and support real decisions without a massive software project on day one. The first deployable layer can be a controlled, domain-specific assistant — which is precisely what lowers the barrier for the industrial teams who need it most.
Back to that night
The point was never to replace the person who retired. People run cement plants; they always will. The point is to make sure that the next time a reading drifts past midnight, what that person knew is still in the room — retrievable, structured, and ready when the decision actually matters. That is the work. And it is exactly why it should be built by someone still inside the industry, not watching from the outside.