A short tour of the unified AI design pipeline — from 1D sizing to 3D manufacturable geometry.
Traditional aerodynamic shape-optimisation workflows are fragmented — separate tools for CAD, meshing, flow solution, and sensitivity estimation, each handoff breaking the gradient chain.
Our platform unifies 1D meanline sizing through 3D CFD-optimised, manufacturable geometry — a capability no existing commercial platform offers.
Potential markets: Gas-turbine OEMs · Turbocharger OEMs · Aerospace Defence & Space · Oil & gas industry · UAV & drone industry · Academic research labs.
One orchestrator, three stages — a single unbroken gradient chain.
Natural-language control via MCP — build workflows, set objectives, analyse results. Multilingual.
1D meanline sizing for radial & axial compressors and turbines — the pipeline's entry point.
Manufacturable blade CAD from meanline parameters, with exact surface sensitivities.
3D RANS optimisation with a Newton-Krylov adjoint solver — exact gradients.
RADAE → Differentiable CAD Kernel → AnamikaCFD, orchestrated by the LLM layer.
Purpose-built GPU hardware — patent pending — so you can run our solvers at full scale, entirely inside your own facility.
Specialised GPU hardware we designed for the platform, tuned to run RADAE, the differentiable CAD kernel, and AnamikaCFD at peak throughput.
Run every solver on-premises. Your geometries, meshes, and design IP never leave your facility — no cloud round-trips, no third-party exposure.
Massively parallel, MPI/GPU-accelerated compute — industrial-scale optimisation and adaptive meshing without leaving your own racks.
The full pipeline runs in-house, at scale, and under your control — no data ever leaves the premises.
One pass yields the full gradient — at a cost independent of the number of design variables — making optimisation over thousands of shape degrees of freedom tractable.