goatai.io/research
Engineering notes from inside the plant
The most durable thing GOATAI can do is explain the hard industrial-AI problems it actually works on — in the open, for the engineers who run physical systems. These tracks are filling in as the engineering matures.
Published
The definition we build on
The category GOATAI works in, defined in the open — the essay the rest of the site answers to.
Track 01
Engineering Notes
Working notes on hard industrial-AI problems — perception under plant conditions, sensor fusion, and the systems decisions behind each product.
Engineering Notes
In preparationA recognition-based perception map for crane clearance
Treating anti-collision as dynamic clearance management — learning a bay’s normal state and detecting novelty, rather than enumerating collision cases.
Draft in preparation
Track 02
Architecture Papers
Deeper write-ups on the reasoning architecture and its physics layers — surrogate modeling, reduced-order physics, and the cycle that stays constant across domains.
Architecture Papers
In preparationSurrogate modeling for BOF and continuous-casting reasoning
Coupling physics reduced-order models with ML surrogates for endpoint and quality reasoning over the converter and caster.
Draft in preparation
Track 03
Technical Essays
Position pieces on why physics-grounded reasoning generalizes where black-box models do not, written for engineers who run physical systems.
Technical Essays
In preparationThe architecture is constant; the physics is grounded
Why one reasoning cycle can run crane, furnace, caster, and ladle — and why that is an engineering claim, not a marketing one.
Draft in preparation
Track 04
Field Observations
What the plant actually looks like up close — site surveys, bay geometry, and the constraints that shape real deployments.
Field Observations
In preparationInside a ladle bay: 36 stands, three cranes, and what coverage requires
Field notes from the DE-bay survey — why per-stand visibility is a geometry problem before it is a vision problem.
Draft in preparation
Collaborate
Working on the same problems?
We share early drafts and methodology with research partners and operating teams. If you run a plant, a shop floor, or a lab working on physics-grounded reasoning, get in touch.