Documentation · v0.3.0-platform-mvp

cufemlab documentation

cufemlab™ is a proprietary GPU-native multi-physics platform from Secrotec B.V., delivered as a hosted SaaS at cufemlab.secrotec.nl. These docs cover sign-up, the public API, the integrated workflows, recipes for each physics domain, and the authoritative scope statement.

Current release: v0.3.0-platform-mvp · 24-phase platform · 290 benchmark checks · 6 GPU-evidence suites · 53 cited materials. Sign up free →

Start here

Getting started

Sign up, claim your 2 free trial credits, submit your first analysis, read the verdict. Five minutes from new account to PASS.

Open the guide →

Motor workflow

The integrated magneto-thermal-mechanical pre-analysis — flux density, slot temperature, rotor safety factor in a single call.

See the workflow →

Examples & recipes

Ready-to-clone job specs for each of the 12 physics domains. Magnetics, thermal, mechanics, fluids, electrochemistry, acoustic, chemistry, optimization, UQ, porous media.

Browse recipes →

API reference

Full REST surface: jobs, results, credits, API keys. Plus the Python cufemlab module namespace if you have on-prem access.

Open reference →

What cufemlab is

cufemlab is a GPU-native finite-element platform that solves magnetics, thermal, mechanics, fluids, electrochemistry, acoustics, multi-body dynamics, plasma, chemistry, optimization and uncertainty quantification on a single CUDA pipeline. Every result cites a published reference, declares its tolerance, reports its measured error, and carries a SHA-256 provenance hash.

It is designed for:

  • Electric-motor and drivetrain R&D teams (magnetics + thermal + mechanics, integrated).
  • EV battery teams (electrochemistry + thermal + pack-level coupling).
  • Power-electronics teams (IGBT cooling, inductors, transformers).
  • Aerospace thermal teams (cooling loops, heat exchangers, compressible gas).
  • Research labs that compare numerical methods and need verdicts they can cite.
  • Coursework and method comparison where reproducibility matters.

Honest scope

cufemlab v0.3.0 is a focused MVP. 18 of 24 roadmap phases are at least PARTIAL today; 6 phases (M magneto-mechanical coupling, S ML/PINN, U CAD import, V adaptive meshing, W multi-GPU/HPC, plus parts of X certification) are on the public 24-phase roadmap with dated estimates. Read the hard limitations page before relying on cufemlab in production work — every line is authoritative.

Discipline

Every result in cufemlab follows ten platform rules — most importantly: no PASS without a cited benchmark, a declared tolerance, and a measured error. The integration runner exits non-zero on any FAIL, and every report carries a SHA-256 provenance hash. See the evidence table on the homepage for the full suite list.

How you access cufemlab

cufemlab is delivered as a hosted SaaS — no installation, no Docker stack, no GPU node to provision. You don't need to manage CUDA versions, Compose files, or NVIDIA Container Toolkit. Three engagement modes:

  • Free SaaS — sign up, get 2 trial credits, run analyses in your browser.
  • Paid plans — Starter / Pro / Business, billed monthly via Stripe. Credits are non-expiring.
  • Pilot or on-prem — for regulated industries or sensitive geometry; we deploy on a GPU node we provision (cloud) or your workstation (on-prem). See pilot tiers.

License & ownership

cufemlab is PROPRIETARY. Copyright © 2026 Eba Turan, all rights reserved. Any use, copy, modification, or distribution requires explicit written permission from Secrotec B.V.

To request a license, evaluation, or pilot, write to info@secrotec.nl.