About the project and the client#
The bottleneck is engineering time. Reading drawings by hand, planning the cutting layout, keeping track of steel grades, working out cost per kilogram — project after project. Sometimes a pile of scrap that can't be reused on later projects.
Before we start coding, we spend several days on site — we walk the shop floor and talk through the process with Mateusz Ozga, a specialist in road-column production. We get the problem from the inside — that's why we can build tailor-made software.
One tool for the whole column production cycle — automated cutting-layout calculation, steel waste minimization through optimal cutting, faster production and a warehouse kept split per project. The real final cost per kilogram and per unit — worked out from real paperwork.
What we're tackling#
The largest cost wasn't steel — it was the risk of human error under a pile of orders and detailed specs. To change that, we have to take on a complex, multi-stage process where every millimeter counts: how to work out the cutting layout with a guarantee of optimality, how to get usable offcuts back into use, and how to land on the real product cost — per kilogram and per unit.
A cutting layout with a guarantee of optimality
Cutting-layout optimization is a combinatorial problem — the math minimum of total bar length and sheet count is out of reach without the right algorithm. The challenge: get an optimal plan with a proof that it can't be beaten, all while handling kerf, stock length limits and the "2 elements from one segment" mode.
The hard rule of never mixing steel grades
Steel grades (e.g. S235 and S355) must never end up together in a single order line item. The system keeps them separate at every level — from the individual plate, through the column, to the bill for the plasma cutter and the invoice.
Usable offcuts back into circulation
Cutting remnants above the minimum required length aren't waste — they go back to a separate offcut warehouse and can cover future demand. The optimization had to tell apart usable offcuts (feeding the warehouse) from shorter ones (counted as loss), and the warehouse had to run separately per construction site.
Module coherence at every stage
Demand, cutting, warehouse, production, costs and invoice — each module pulls data from the previous one, no manual re-entry. The challenge: hand the result of one stage to the next without losses, so the whole path from construction site to invoice is one coherent process.
Got a similarly complex project? Let's talk.
Provable optimization, a tough production domain and real cost accounting — we know how to pull it together into one coherent system.
How we approach the project#
Three principles that guide this project: a deep understanding of the process first, then code — long, detailed talks on site at the production plant. Well-thought-out algorithms validated on real projects — algorithms designed for this specific process and tested against the client's real paperwork. Simplicity of use and integration with the existing production process — one tool for the whole process, with exports going straight where they're needed.
Understanding the process at the production site
We start with days spent at the client — walking the workshop, talking to operators and engineers, watching the process from site demand to invoice. We only start building the data model and cutting rules once we get how the product is actually made.
Well-thought-out algorithms validated on real projects
We design the 1D and 2D cutting algorithms for this specific process, and we test each one on real projects and the client's real production documents. A solution only goes live once we've proved it beats manual calculations.
Simplicity of use and integration with the existing production process
One tool tying together demand, warehouse, cutting, production and accounting — no switching between systems. The interface walks the operator through every stage, and the result goes straight where it's needed: a DXF export to the plasma cutter, a delivery note to the galvanizing plant, an invoice through Connecto and a full project document pack to download (PDF, XML, CSV).
App views#
The full flow from the dashboard, through demand from workshop drawings and cutting optimization, the warehouse, production and costs, to Connecto invoicing and the completed project archive.
Some app screens
Views from the working system — the full cycle from the dashboard, through demand, material and cutting optimization, to production, costs and e-invoices.
What we deliver#
Biggest wins after the rollout — measured in engineering time, amount of waste, number of errors and coherence of the whole column production process.
Decisions that made the difference#
Four decisions that mattered most for the outcome — matching solvers to the problem type, a hard separation of steel grades, splitting long plates into welds and a CAD-compliant export. A web app: React + NestJS, a standard tech stack with no exotic dependencies.
A separate algorithm for 1D cutting and for 2D nesting
Bar cutting is a 1D problem — a MILP solver (HiGHS) delivers a plan with a proof of optimality. Plate nesting is a 2D problem — a layout of differently sized plates (maxrects) in thickness×grade groups plus analytical sheet selection for identical plates.
Two fitted algorithms give a better result than one universal heuristic.
A hard separation of steel grades across the whole process
The steel grade (S235 / S355) is tracked separately at every level — from the individual plate, through the column, to the bill for the plasma cutter and the invoice. Two grades will never end up in a single order line item.
The rule holds throughout the process, because a grade mistake is a real production defect, not a cosmetic one.
Plates longer than a sheet split into segments automatically
Long plates are split into welded segments with an automatic report of the number of welds and welding schematics. Plates with mounting holes are never split.
This makes the optimization result match what can actually be produced, rather than a fantasy that assumes infinite sheets.
DXF ready for CAM and Connecto integration
The cutting layout is exported to DXF in a format compatible with plasma cutters and CNC machines — the file goes straight to the machine, no manual rework, no conversion.
We don't build the accounting from scratch. Full Connecto integration via the API closes the process from site demand to the invoice.
Technical core#
Tools chosen for this process — proven algorithms and formats, no exotic dependencies.