AI that runs on your own infrastructure, trained on your own data: agents that take over repetitive engineering work, surrogate models that predict in minutes, and a digital thread across your engineering and production data.
01What we deliver
From a first workshop to models in production. Start wherever your team is, and take only the paths you need.
AgentCrew · Spectra · Iso
Prism
Sift
02From the Lab
Method-first technical writeups of studies run on our CAE/CFD and surrogate-model pipelines: what we ran, how it was verified, and where each method reaches its limits.
2,560 propeller designs screened with fast blade-element theory, then corrected with 12 blade-resolved CFD runs, instead of running CFD on every candidate.
A post-processing sign error slipped into an automated study. A verification agent flagged it as physically implausible before it reached the report.
A surrogate trained on high-fidelity thermal runs predicts component temperatures in milliseconds. The hard part is knowing exactly where to trust it.
03WHY CURLSCAPE
04How we engage
It starts with a call, then a fixed-price sprint, so you can prove the approach before committing to a build. No open-ended retainers to start, no rip-and-replace.
Intro call · 2–4 week sprint · 8–16 week build
The discovery sprint is fixed-price and takes one specific engineering problem, so you see the approach work on your own data before you commit to a build. A new connector to your stack usually takes days to weeks, not months.
Your IT and security requirements set the architecture. We deploy on your cloud or on your own servers, including fully air-gapped, so design data stays in your environment.
Your simulation archives, CAD and test data train the models built for you. Ownership of what the pipeline produces is yours, written into the engagement.
You keep your tools; we automate the work between them. Our agents drive Siemens NX, Solid Edge, Ansys and OpenFOAM through their own interfaces, and Prism works from your PLM, ERP and test exports.
We automate setup, data transfer and routine checks, and you decide where a person signs off. In Spectra, nothing is committed to PDM without an engineer's approval, and a design-rule violation stops the design rather than raising a warning.
It starts with a call and ends with a system your team owns. Paid engagements are scoped and costed upfront, so you know the number before you commit.
05FAQ
Curlscape helps enterprises build sovereign AI: AI systems that run on their own infrastructure, trained on their own data. It works across three areas: agentic AI that takes over repetitive engineering and operations work, surrogate models that learn from past simulation and test data and predict results in minutes instead of days, and a digital thread that connects engineering, production and business data, so AI agents can find patterns no single system shows. Teams not sure where to start can begin with a workshop or with training.
See the five pathsCurlscape offers five products. Curlscape AgentCrew runs AI agents that prepare geometry, mesh, set up solvers and write reports. Curlscape Spectra turns specs and RFQs into rule-checked CAD models through repeatable design workflows. Curlscape Iso is an AI assistant that works inside SolidWorks and Solid Edge. Curlscape Prism answers questions across engineering, quality and supply-chain data and shows the records behind each answer. Curlscape Sift builds surrogate models that predict simulation results in minutes.
Compare the productsCurlscape builds multi-fidelity surrogate models for engineering simulation, fusing many cheap low-fidelity runs with a few expensive high-fidelity ones using co-Kriging. In a published 1U-server thermal study, 10 fine-mesh CFD runs corrected a coarse-mesh model and cut cross-validated error by 62–67%. In a propeller study, 2,560 blade-element evaluations fused with 12 CFD runs cut error by 95% for thrust and 91% for torque. Both studies state their limits.
Read the multi-fidelity studiesCurlscape AgentCrew runs AI agents that handle the repetitive layers of CAE work: planning, geometry preparation, meshing, solver setup and reporting. Each agent drives the tool through its own scripting and API layer, and review agents check each stage before the pipeline moves on. Current connectors cover Siemens NX, Solid Edge and FreeCAD for CAD, Fluent Meshing, Ansys Meshing and snappyHexMesh for meshing, and the Ansys solver suite and OpenFOAM for solving.
How AgentCrew worksA surrogate model is a machine-learning model trained on past simulation results that predicts new results in minutes instead of hours or weeks. Its accuracy holds inside the range of designs it was trained on, so Curlscape treats that trained envelope as part of the deliverable and verifies key designs with a full simulation. Curlscape Sift builds neural operator surrogates (DoMINO, AB-UPT, MeshGraphNet), ML-based model order reduction and multi-fidelity models.
Where to trust a fast modelYes. Curlscape deploys on your cloud or on your own servers, including fully air-gapped, so geometry, BOMs and test data stay in your environment. Curlscape Spectra, for example, can be hosted at your site and run without any outside connection.
Security and IPYou do. Everything runs on your infrastructure, so your geometry, BOMs, and field data stay in your environment. IP ownership of what the pipeline produces is yours, written into the engagement.
Security and IPStart with one specific engineering problem. Curlscape begins with a call, then a fixed-price discovery sprint of 2–4 weeks that tests the approach on your own data and ends with a costed build plan; a full build typically takes 8–16 weeks. Teams not yet sure where AI fits can start with a workshop that produces a concrete deployment plan.
Schedule a call
