Web Studio
Strategy, product and planning layer (internal). The customer-facing Studio shows only Overview + Run.
Internal planning cockpit — admin only. Valuation, roadmap, analyzers and risk below are internal CEO/strategy planning data, not customer-facing product claims.
Open Studio Run →
Planning layer. Valuation figures here are conditional planning assumptions for CEO review, not valuations Cufemlab holds. Cufemlab today is a verified GPU FEM solver + API, not COMSOL/Ansys/Altair. FNO is an in-distribution surrogate / self-validating demo, not a general-purpose neural operator.
CUFEMLAB · STUDIO
Electric-machine analysis workspace
Run verified, GPU-assisted electric-machine analyses in the browser or via the API. Every run returns result values, an independent validation card, the exact execution backend, raw artifacts and a generated evidence PDF — so you can trust the number, not just read it.
21 workflows
15 self-service + 6 machine live-beta
GPU-assisted · verified · evidence-first
How a run works
Step 1 · RunPick a workflowChoose a verified analysis type, set parameters or keep the defaults, and submit. Most run from parameters — no file upload.
Step 2 · ResultRead the valuesThe result page shows headline metrics and every scalar value the solver produced.
Step 3 · EvidenceCheck the proofValidation passport (independent cross-checks), execution backend (GPU/CPU), raw JSON and downloadable artifacts.
Step 4 · ReportGenerate a PDFTurn a completed job into a generated PDF evidence report (SHA-256 tamper-evidence) you can share.
Loading…
Current product state
Positioning
Next:
Catalog workflows
Listed in the API + UI picker while the internal-beta flag is on
Self-service / live beta
+
always listed · listed while the internal-beta flag is on · internal-only, never listed
Catalog status
Aligned
All workflows: the API listing (internal-beta flag on), the app picker and the generated catalogs agree · enforced by test_f13_catalog_truth.py
Honest limitations
What every run gives you
Result valuesHeadline metrics plus every scalar the solver produced, on the result page.
Validation passportIndependent cross-checks (FEMM / GetDP / analytic) with reference, error % and tolerance where available.
Execution backendHonest GPU vs CPU: what actually ran, and whether GPU was required — a measurement, not a label.
Raw JSON + artifactsThe full result JSON and downloadable files. Nothing hidden.
Generated PDF reportOne-click evidence PDF with a SHA-256 tamper-evidence digest (not a PKI signature).
6 machine-analysis workflows are live-beta. They are internally verified (oracle-closed vs GetDP / FEMM / analytic) and GPU-verified, but not enterprise/commercial-certified and carry no external validation passport yet. The 15 self-service workflows — and the flagship moving-band cogging torque, cross-checked vs FEMM & GetDP — are the most mature. Cufemlab is a pre-analysis and evidence layer, not a replacement for Ansys / COMSOL / JMAG / Altair.
Roadmap phases
Analyzer snapshot
| Analyzer | Score | Summary |
|---|---|---|