Skip to content
Skip the journey

Every great design starts with a question

Nabla AI is developing a computational fluid dynamics (CFD) simulation engine to help engineers test ideas with less setup.

Discover Nabla

A new perspective on CFD

The experience our founders bring to Nabla

  • Imperial College London
  • CFIS · UPC
  • NASA
  • Rolls-Royce
  • CIMNE
  • Barcelona Supercomputing Center

Engineering is hard
The workflow shouldn’t be

Manual setup. Long waits. Uncertain results. Too much work between a design and an answer.

A conventional CFD workflow
  1. Prepare
    Configure the case
  2. Mesh
    Resolve the geometry
  3. Run
    Wait for convergence
  4. Interpret
    Reconcile the results
Failure can send you back to setup.
01

Days in preparation

Setup and manual meshing delay the first run.

02

Time lost to restarts

A failed simulation can waste days of compute.

03

Same design. Different answers

Different teams’ setup choices can change the result.

The cost is more than compute

With conventional CFD, faster iteration can mean lower resolution. Building confidence in the results takes refinement and repeat runs. This means more time, more compute and more pressure on engineers.

Less friction
More understanding

We’re developing a new CFD engine to make physical simulation easier to use and understand.

A clearer view of the physics

A defined geometryDetail you can examine
Assumptions

The conditions behind the study

Flow detail

The behaviour around the design

Checks

The basis for confidence

Conceptual study of a wing section. Geometry and flow are illustrative, not simulation results.

  1. Bring the design. Skip the manual mesh

    Automated preparation, without conventional, geometry-fitted volumetric meshing.

  2. Let the physics guide the computation

    Resolution concentrated where the flow requires it, with physical simulation at the core.

  3. Turn results into a clear report

    Detailed reports generated from the results, tailored to the project, the client and the questions that matter.

What we’re working towards

  • Faster simulations
  • Lower computing costs
  • Higher resolution
Talk about your project

How the approaches compare

From preparing a design to sharing the findings. Where established tools focus, and what we’re building at Nabla.

Scroll to compare

Nabla’s intended workflow compared with established CFD and AI surrogate approaches
The workflowNabla AIIn developmentEstablished CFDAnsys Fluent · STAR-CCM+ · OpenFOAMAI surrogate modelsPhysicsX · Neural Concept
Preparing the geometryAutomated preparation without a conventional volume mesh.Volume meshing and solver setup, with automation options.Geometry inputs prepared for a trained or pre-trained model.
Physical basisA new adaptive CFD engine, with physical simulation at its core.Numerical solution of the governing physical equations.Predictions learned from simulation or experimental data.
Level of detailResolution concentrated where the flow needs it.Mesh refinement and numerical settings, including adaptive options.Detail depends on the training data and model architecture.
Confidence in the resultPhysical results accompanied by assumptions, checks and limitations.Convergence, mesh sensitivity and validation checks.Model validation, applicability checks and uncertainty assessment.
Sharing the findingsDetailed reports generated for the project and client.Post-processing and reporting tools, configured for the study.Predicted fields and performance metrics in engineering applications.

Nabla’s column describes our intended workflow. Established products can combine these approaches; capabilities vary by tool and configuration.

Sources and scope

Workflow comparison based on public product documentation, reviewed September 2026. This is not a performance benchmark.

What are you
working towards?

Let’s start a conversation

We would love to hear from engineering teams, researchers and investors.