Drop in a geometry, run, read the results. No meshing, no parameter tuning. Nabla AI is building a CFD engine that is faster, cheaper, more accurate and supercharged by AI.
Every aircraft, ship, submarine and rocket is a shape that has to survive contact with air or water. CFD is how engineers find out on a computer whether it flies, goes fast or holds together, before anyone cuts metal.
Yet the workflow has not fundamentally changed in three decades: wrap the geometry in a body-fitted mesh, tune the parameters, hand it to a finite-volume solver, wait. The industry feels that cost every day.
The run diverges on day three.
You spent a week on the mesh, handed it to the solver and waited. Then a handful of bad cells blew up the residuals, and you are back at the mesh. Again.
Preprocessing eats the schedule.
NASA's CFD Vision 2030 study names mesh generation as a major workflow bottleneck, often dominating the human effort a simulation requires. The physics has not even started.
Same geometry, two engineers, two drag numbers.
The answer depends on the mesh and the parameters you picked, not only on the physics. Nobody actually wants that.
It takes a whole team.
Meshing specialists, solver experts, an HPC budget and weeks of calendar time. All of it for a single design point.
The wind tunnel is still the fallback.
Not because blowing air at a scale model is cheap, fast or convenient, but because nobody has made simulation trustworthy enough to replace it.
Turbulence gets approximated, not resolved.
Quieter, safer, more efficient vehicles are decided by eddies a tenth of a hair wide. Resolving them costs more than anyone can afford, so it rarely happens.
Nature does not build a mesh or tune a model: the flow resolves itself. We have derived the formulation that makes this possible in an engineering tool, the one whose absence has held this class of methods back for thirty years, and validated it analytically and in simulation.
Straight from CAD. No cleanup, no body-fitted volume mesh, no parameters to tune.
Resolution follows the physics. Compute concentrates where the flow demands it, not where you guessed it would.
One answer per geometry, independent of who set it up. Analyse, iterate, run the next candidate.
Quieter, safer and more efficient aircraft are decided by turbulence, and turbulence has to be resolved, not approximated. Engines, pylons, flaps and wingtips are exactly where body-fitted meshes break. With Nabla the geometry goes in as it is, and resolution concentrates on the boundary layer and the wake, where drag and noise are actually decided.
Hull resistance and propeller wakes are long transient simulations, the kind where a mesh failure on day three hurts most. Without a mesh to fail, a hull form is evaluated from the STL, with resolution concentrated on the free surface and the wake instead of spread uniformly across the domain.
Geometries for drones, vehicles and rotor blades are now generated straight from performance objectives, faster than anyone can validate them. Each candidate used to need its own mesh. With Nabla each one is just another STL, evaluated with real physics rather than a surrogate that returns a plausible answer with no guarantee it holds for a design nobody has simulated before.
The noise and wake signature of a submarine are decided by the smallest eddies around the hull and sail. Resolving them below a tenth of the width of a human hair, the scale at which they dissipate, is the resolution the industry needs and that nothing on the market delivers at an affordable cost.
Three ways to find out how a shape behaves in a fluid today, and what changes when the mesh goes away.
As AI generates more candidate designs, our ambition is to make physical evaluation fast enough to support a continuous cycle of generating, simulating and improving them.
Nabla AI is building the computational layer to help make this possible.
Aerospace engineers and computational physicists who have built and run CFD at:
We are talking to engineering teams tired of meshing, simulation companies, researchers and investors interested in the next generation of CFD. If that is you, we would like to hear what you are running today.
Barcelona · San Francisco