LoopForgeLab

A design-stage cost estimation system for machine builders. It answers the two questions every business must answer for an order before the design is finished — what will this cost to make, and what will it cost the customer to run (TCO), and how can I improve this design. It does this from the data the design engineer already has.

What we do?

Takes in a BOM export from ERP, PLM or CAD, or a STEP assembly. The engine deduplicates parts on identity and geometry, checks units and masses for plausibility, classifies every line as make or buy and, for made parts, by process family (machined, sheet-metal, welded, cast, turned), and returns a cost breakdown per part and per assembly: material, process, setup, overhead, purchase price. Tolerances are modelled per feature, not per part. Forge engine uses reasoning on 3D, BOM and field data the company owns and does not own, Moroever the engine provides recommendations on materials, tolerances and geometry. 

Every figure carries two annotations: its source (engineer declaration, named rule, catalogue lookup, model estimate) and its confidence. Where the engine cannot classify a part reliably it says so and asks, with the open items ranked by cost impact, rather than printing a confident number that turns out to be a guess. Resolving the open items on a 100-part subassembly typically takes an engineer well under an hour.

A second module projects multi-year cost of ownership — energy, maintenance, spares, downtime, labour, end of life — from the same design data and a set of operating drivers, with sensitivity analysis on the drivers that matter. Sales engineers can put a defensible TCO appendix on a quote instead of a sales narrative; design engineers can see the operating-cost consequences of a drive or bearing choice while it can still be changed. The lifecycle engine runs on declared drivers today; calibration against a customer's own field data (MTBF, energy profiles, spares consumption) is the next development phase.

Because both modules share one data model, the engine compares design variants: welded versus bolted frame, custom versus catalogue linear axis, aluminium versus steel cover. Deltas cancel shared rate-card errors, so comparisons are reliable from the first week of a pilot, before any calibration.

Who is it for?

Engineer-to-order machine builders where the price is committed before the design is complete: packaging and material-handling OEMs, special-purpose machine builders, measuring and test-system makers, assembly automation, fixture and gauge builders, vacuum and coating equipment. Typically 30 to several thousand employees, headquartered in DE/AT/CH/PL/CZ/NL/BE. Companies with a high volume of smaller projects are especially well served, because quoting time is the measurable saving.

 Data

EU-hosted, single-tenant. No cross-customer pooling. Geometry, drawings and prices are never used to train models that serve other customers. Data-use terms are agreed in writing before the first file is sent.

Who are we?

Founded by an engineering team with 15+ years of production software and AI experience and a PhD in AI; developed through the Carbon13 venture programme; validated in 70+ structured interviews with design engineers, industrial engineers, buyers and managing directors across the machinery sector. Currently in pilot with a large European packaging-machinery OEM and a second machine builder.

What are we looking for?

Two to three further pilot partners willing to test the engine on two to five completed projects with known actual costs. You provide BOM exports (STEP files, drawings and rate cards welcome, not required); we return the engine's estimate, a comparison against actuals with every deviation explained, and a calibrated model of your part mix and rates that stays yours. Effort on your side: a few engineer-hours over two weeks. Success is measured on estimate deviation, time to first usable estimate, decisions changed in design review, and whether your engineer would use it on the next enquiry unprompted. We also welcome OEMs with an installed base who can share anonymised service data, and research partners for EU-funded work on predicting operating outcomes from design parameters under sparse field data.