Designing underfloor heating is skilled but repetitive work: trace the floor plan, size the loops against heat-loss tables, balance the manifold by trial and error. Our client, an underfloor heating specialist working to Dutch and European standards, wanted that turned into software so their engineers could take on more projects without hiring in proportion. We built the platform end to end. Computer vision reads an uploaded plan and extracts rooms, walls, doors, and windows; a geometry engine generates continuous bifilar pipe spirals that respect PE-X bend limits; and a deterministic optimizer balances every loop, returning the flow rates and valve presets an installer needs on site.
View Impact
The client's engineers were spending days per project on drafting and calculation that followed the same rules every time. We collapsed it into one workflow: upload a plan, calibrate the scale, place the manifold, and get a validated layout, with the physics still traceable to the standards their work is audited against.
A pipe layout is only useful if it can actually be installed. The system had to fill irregular room polygons with a continuous, non-intersecting spiral, never bend tighter than PE-X allows, keep water velocity in a safe band, and still satisfy each room's heat-loss target.
We split physics from geometry. A C# and C++23 service runs the certified ISSO 51, 53, and 57 heat-loss kernels, while a FastAPI backend handles vision and routing — YOLO and PaddleOCR for plan extraction, Shapely for millimetre-accurate model space, and a custom spiral generator using smoothstep interpolation and Chaikin smoothing.
An interactive Nuxt 3 canvas where engineers drop manifolds and watch layouts regenerate live, with per-circuit pipe length, velocity, pressure drop, and valve preset values. The optimizer finds the lowest feasible supply temperature, cutting energy use while holding surface temperatures within comfort limits.
View Impact
The client's engineers were spending days per project on drafting and calculation that followed the same rules every time. We collapsed it into one workflow: upload a plan, calibrate the scale, place the manifold, and get a validated layout, with the physics still traceable to the standards their work is audited against.
A pipe layout is only useful if it can actually be installed. The system had to fill irregular room polygons with a continuous, non-intersecting spiral, never bend tighter than PE-X allows, keep water velocity in a safe band, and still satisfy each room's heat-loss target.
We split physics from geometry. A C# and C++23 service runs the certified ISSO 51, 53, and 57 heat-loss kernels, while a FastAPI backend handles vision and routing — YOLO and PaddleOCR for plan extraction, Shapely for millimetre-accurate model space, and a custom spiral generator using smoothstep interpolation and Chaikin smoothing.
An interactive Nuxt 3 canvas where engineers drop manifolds and watch layouts regenerate live, with per-circuit pipe length, velocity, pressure drop, and valve preset values. The optimizer finds the lowest feasible supply temperature, cutting energy use while holding surface temperatures within comfort limits.
Forging Digital Solutions with precision and passion. From freelance roots to global impact.
Trusted & Verified
