Smart Steel Technologies

The problem

Steel yards accumulate material that nobody ordered: overproduction, and coils that came out of the mill slightly wrong. It occupies space, costs money to store, and usually ends up sold at a discount or scrapped.

Some of it could fulfil a new order instead — but working out which piece, for which order, through which production path is genuinely hard. A slab can become hot-rolled coil, then cold-rolled, then galvanised, and each mill has its own permitted paths, constraints, priorities and rules. Planning departments were doing that reasoning by hand.

What I built

I spearheaded the Material Allocation product — Order-to-Material — with technical ownership from the initial data modelling through to the final user interface.

That meant architecting the domain: technical orders, production orders, material substructures and KWW positions. Then the core kernel scoring and enhancement logic that ranks candidate matches, the flow that creates and publishes an allocation, and end-to-end integration tests to keep the whole path stable.

And the parts planners actually touch: allocator views, free allocation, autocompletion endpoints, saved search filters, and Excel reporting fast enough to be usable on real volumes.

The result

A tool that produces ranked, automated reallocation proposals instead of a person reasoning through production paths by hand — deployable standalone or alongside the company's short-term and long-product schedulers.

Two years and four months in a steel-industry scheduling department also taught me the thing no greenfield project can: that in heavy industry the software is judged against what the planners were already doing, and being right is not enough if it is slower.

Stack

Services this involved