01
Aerospace
Aerodynamics, propulsion, hypersonics and complex turbulent flows.
Physics-first · GPU-native
High-fidelity CFD, rebuilt for the GPU era.
Nabla AI is building a physics-first, GPU-native simulation engine that dynamically concentrates compute where it matters most.
- Dynamic resolution
- Solver orchestration
- LES → near-DNS fidelity
Adaptive simulation previewAMR · L6
vorticity∇ · u = 0
Live in-browser demo — cells refine around the wake as vortices shed.
The problem
High-fidelity simulation remains too slow and too manual.
A single trustworthy result can take days or weeks — most of it spent preparing the simulation rather than learning from it. Today's high-fidelity CFD workflows still depend on:
- Manual meshing and geometry clean-up
- Hand-tuned models, solvers and parameters
- Uniform or poorly allocated resolution
- Long iteration cycles
- Expensive compute infrastructure
- Highly specialised engineering teams
Traditional CFD workflow
repeat for days–weeks
- Geometry
- Meshing
- Solver setup
- Run
- Debug
- Remesh
- Run again
Nabla AI workflow
automated · adaptive · continuous
- Geometry
- Physics intent
- Adaptive simulation
- Results
Technology
A simulation engine that understands where compute matters.
Resolution is treated as a dynamic resource — allocated in space and time by the physics of the flow, not fixed upfront by a mesh.
01
Dynamic resolution allocation
The engine continuously reallocates computational resolution toward shocks, vortices, boundary layers and other physically relevant structures — instead of spending compute uniformly across the domain.
02
Automated solver orchestration
Nabla AI selects and coordinates numerical methods, fidelity levels and solver configurations based on the evolving flow — no manual tuning loop.
03
GPU-native architecture
The complete simulation pipeline is designed around modern GPU hardware, rather than adapted from legacy CPU-based architectures.
04
Configurable fidelity
Users can balance speed and accuracy across a spectrum ranging from engineering-grade LES to near-DNS fidelity, per case and per question.
Approach
Not another CFD surrogate model.
Neural surrogates trained on libraries of precomputed simulations can be extremely fast — but they inherit the limits of their training data and offer no physical guarantees on genuinely new designs. Nabla AI is a numerical simulation engine first; machine learning enters only where it accelerates the physics without compromising it.
- Physics-first numerical simulation
- Adaptive computation
- GPU-native execution
- Machine learning where it creates measurable value
- Physically consistent outputs
- Generalisation beyond narrow training distributions
Traditional CFD
- Physically rigorous
- Slow and manual
- Expensive
- Complex workflows
Pure AI surrogate
- Extremely fast
- Limited generalisation
- Training-data dependent
- Hard to trust
Nabla AI
our approach- Physics-first
- Adaptive compute
- GPU-native
- Automated workflow
Applications
One engine, multiple engineering domains.
The same adaptive core applies wherever turbulence, heat and high-speed flow decide how a product performs.
02
Automotive
External aerodynamics, thermal management and design optimisation.
03
Energy
Wind, combustion, heat transfer and next-generation energy systems.
04
Turbomachinery
Compressors, turbines and rotating flows requiring high-resolution simulation.
05
Defence
High-speed flows, propulsion, autonomous systems and complex mission environments.
06
Industrial flows
Multiphysics and turbulent flows across advanced manufacturing and infrastructure.
Vision
Simulation should become an interactive engineering tool.
Instead of waiting days for a simulation result, engineers should be able to explore designs, test hypotheses and iterate at the speed of thought.
Nabla AI is building the computational layer that makes this possible.
Company
Built by engineers working across CFD, aerospace and computational physics.
Jesús Navas Guerrero
Co-founder · CEO
- Aerospace Engineering and Engineering Physics
- MSc in Advanced Mechanical Engineering, Imperial College London
- Computational fluid dynamics, plasma physics and computational MHD
- Former researcher at NASA Goddard
- La Caixa Fellow
Martí Massó Moreno
Co-founder · CTO
- Aerospace Engineering and Engineering Physics
- MSc in Astrophysics, Particle Physics and Cosmology
- Former CFD engineer at Near Space Labs
- Research experience at NASA Goddard
- Computational models for ICME reconstruction and propagation
Contact
Let's build the future of engineering simulation.
We are currently speaking with engineering teams, simulation companies, researchers and investors interested in the next generation of CFD infrastructure.
Barcelona · San Francisco