Independent research lab · Compressible & reactive flow
Shocks, detonations, stiff chemistry, and refinement that moves with the front. We derive the methods and write the code.
2.29M cells
Airframe case on one rank
Mixed hex and polyhedral, with dynamic refinement, inside 16 GB.
217B/cell
Resident footprint
Audited line by line. 241 B/cell with a nine-species reacting registry.
10−13
Conservation across an adapt
Mass and energy closure, checked every refinement cycle.
330sites
Standard-C++ parallel loops
No vendor kernel language anywhere in the solver.
An independent research lab
The best methods in compressible and reactive flow are developed inside a handful of large institutions, published on a multi-year cycle, and arrive — if they arrive — a decade later, wrapped in a licence. The bottleneck is not the science. It is the gate in front of it.
We keep what makes research trustworthy — falsifiable claims, canonical verification, honest limitations — and run it on an engineering clock.
Accuracy over aesthetics
Our field sells itself with colour. We optimise for the number being right. The flow above is a live solver, not a video — move across it and read the velocity and pressure it is actually carrying at that point. A computation you can interrogate is evidence; a rendering is not.
Discipline, engineering clock
Falsification conditions written down before testing. Canonical verification and byte-level reproducibility gating every change — against a project schedule, not a publication cycle.
An unusual combination
Deriving a scheme and making it saturate a GPU are normally two people in two departments. Here they are the same job, which is why the numerics and the implementation fit each other.
What we model
Most CFD is comfortable. These are the flows where discretisation, chemistry and mesh must all be right at once — and getting one wrong costs you the physics, not a decimal place.
01
Transonic through hypersonic, on geometry that came out of CAD rather than a tutorial. Unstructured polyhedral meshes, surface-snapped walls, refinement that follows the shock system.
02
Detailed kinetics, stiff sources coupled to the hydrodynamics rather than split from them, and an energy formulation refinement cannot corrupt.
03
Dynamic octree adaptation driven by the solution, as data-parallel passes over space-filling-curve keys. Nothing returns to the host to regrid.
The platform
Three components, one stack. Each solves a problem the field has largely chosen to live with.
Header-only compressible and reactive finite-volume solver. Flux, reconstruction, closure, integrator and chemistry are compile-time types fused into one update, with nothing dispatched at runtime.
DetailA compiler plugin that lowers standard C++ parallel algorithms to SPIR-V, so unmodified source runs on any vendor’s GPU — including hardware current toolchains have dropped.
DetailHermetic, content-addressed build and dependency toolchain. One command, no system toolchain assumed, and a lockfile that regenerates the same binary.
DetailFrom surface to answer
Not a simplified representation of it. A triangulated surface out of CAD goes in; a partitioned, body-fitted, adaptively refined mesh and a solved flow field come out — through a pipeline we built end to end, so that when the physics needs the mesh to change, it can.
Open questions
Held as questions, not products. Each states why it is hard, how we attack it, and what we still cannot determine.
A real shock system can be photographed in flight — but only for about three seconds at a time, from one line of sight, under conditions that almost never line up.
How we approach itCapturing a shock costs artificial dissipation. Spend it in the wrong place and you damp the instability you were trying to predict.
How we approach it
Chemistry sets a length scale orders of magnitude below the device. Under-resolve it and the wave speed is not slightly wrong — it is qualitatively wrong.
How we approach itA threshold phenomenon, geometry-dependent, and the question hydrogen explosion safety turns on. Also where every shortcut in a reactive solver surfaces.
How we approach itHow to work with us
We are looking for industry collaborators, research sponsors and academic partners. Three relationships, one standard of evidence.
Industry collaboration
Taken on as a research problem with an engineering deadline. You get the method, the evidence and the limitations.
Research sponsorship
Results as they are produced, negative ones included, and influence over what gets attacked next.
Academic partnership
Benchmarks, shared validation, independent cross-checks and co-authored work with university groups.
Open for 2026
Tell us what you are trying to predict, and what has gone wrong so far. If it is outside what we do, we will say so and point you at who does it.