Skip to content
Apex
ApexProductsAether for engineering

Aether for engineering.

AI that retires the engineering CAE stack.

Aether for engineering is what the incumbent commercial CAE stack collapses into when the model is the substrate. CFD, FEA, electromagnetics, multiphysics, acoustics, thermal, fatigue, additive manufacturing — unlimited workloads across ten engineering disciplines, in one project file, on one set of weights.

AI for
Every workload
Coverage
Unlimited
Model
One Aether
Validation
Measured reality
Featured workloads

The hardest engineering in the world, under one model.

These are the workloads Aether is built for — the ones that today require a stack of seven-figure tools, a team of specialists, and weeks of mesh-and-pray. One model, one project file, the same answer at the end.

01 · Launch vehicles

Rockets and reusable launchers.

1.8%
drag error vs flight data

Ascent aerodynamics across subsonic, transonic, supersonic and hypersonic regimes. Combustion-chamber CFD with reacting chemistry. Regeneratively-cooled nozzles. Turbopump cyclics. Fairing acoustics. Stage-separation dynamics. Re-entry thermal. One project file, every loadcase your launch programme needs to clear flight-readiness review.

What Aether computes
  • Ascent CFD — RANS, DDES, LES across the full envelope
  • Reacting-flow combustion · regen cooling · nozzle expansion
  • Coupled aero-thermal-structural on TPS and hot structures
  • Pogo, slosh, POGO, fin-flutter, fairing acoustic loads
  • Stage-separation 6-DOF + multi-body contact
  • Trajectory optimisation under wind + thrust uncertainty
02 · Nuclear fusion

Tokamaks, stellarators and Z-pinch.

18 MW/m²
divertor heat flux resolved

Magnetic and inertial confinement under one model. MHD equilibrium, edge plasma transport, divertor heat exhaust, breeding-blanket neutronics, magnet coils at 4 K under fault loads, first-wall thermal under simultaneous neutron and heat flux. Operates with the boundaries that fusion programmes already work inside — tritium handling, export-control, ITAR-clean compartments.

What Aether computes
  • MHD equilibrium · gyrokinetic transport · edge SOL
  • Magnet coil EM + structural at 4 K · quench analysis
  • First-wall thermal under neutron + heat flux coupling
  • Divertor design — pumping, particle handling, heat exhaust
  • Breeding-blanket tritium production · neutronics
  • Disruption mitigation · plasma control scenarios
03 · Automotive

Passenger vehicles, EVs, autonomy.

ΔCd 4.2%
vs wind-tunnel correlated

Drag, lift and cooling on the body. Battery-pack thermal and electrochemistry. Crash on coupled body-in-white and battery modules. NVH from powertrain to cabin. Aerodynamics on rolling wheels and active ride-height. Sensor placement for ADAS — radar cross-section, lidar return, camera glare — under one mesh.

What Aether computes
  • External aero — drag, lift, soiling, side-wind
  • Battery electrochemistry coupled with pack thermal
  • Full-vehicle crash · battery intrusion · TPRD venting
  • NVH — powertrain mount, cabin noise, road excitation
  • Sensor coverage — radar, lidar, camera fusion
  • Underhood thermal · charge-port cooling
04 · Drones & eVTOL

UAVs, eVTOL, urban air mobility.

−4.1 dB
perceived noise · ground effect

Rotor wakes, blade-on-blade interaction, ground effect, gust ingestion. Coupled flight-dynamics with structural flexibility. Acoustic signature for community-noise compliance. Battery thermal under high C-rate. Sensor and link-budget under realistic atmospheric loss. The whole vehicle — propulsion, control, structure, noise — under one solve.

What Aether computes
  • Actuator-disc and full-blade rotor CFD
  • 6-DOF flight dynamics with flex-body airframe
  • Acoustic emission · psychoacoustic metrics
  • Battery pack thermal under sustained high C-rate
  • Gust ingestion · wind-shear loadcases
  • Link-budget + atmospheric attenuation
05 · Robotics

Industrial arms, mobile robots, humanoids.

142 Nm
shoulder torque · 8 kg payload

Trajectory planning with inverse-kinematics under joint-torque limits. Flex-body link FEA so high-acceleration moves don't excite resonances. Contact dynamics with realistic friction. Motor and drive electrothermal. Battery and tether energy budgets. Tested against your real fixtures, your real payloads.

What Aether computes
  • Inverse kinematics under joint torque + reach limits
  • Flex-body link FEA · resonance avoidance
  • Contact-rich manipulation · friction · slip
  • Motor + drive electrothermal envelope
  • Battery and tether energy budgets
  • Whole-task planning with reversible recovery
06 · Hypersonics

Hypersonic vehicles and re-entry.

3.2 MW/m²
stagnation heat flux

Real-gas thermochemistry. Shock-boundary-layer interaction. Coupled aero-thermal-structural on TPS. Inlet/isolator unstart prediction. Material ablation and recession. The physics the incumbent CFD codes treat with linearisations, treated end-to-end.

What Aether computes
  • Real-gas thermochemistry · Park 2T model
  • Shock-BL interaction · turbulence transition
  • TPS material ablation · recession · pyrolysis
  • Coupled aero-thermal-structural at Mach 5–25
  • Inlet/isolator unstart prediction
  • Radiative + conductive heat exchange
07 · Battery & energy

Cells, packs, fast-charge.

54 °C
T_max · 4C fast charge

P2D electrochemistry coupled with thermal at cell, module and pack scale. Fast-charge degradation. Thermal-runaway propagation with vent and TPRD modelling. Mechanical abuse under crash. Solid-state and silicon-anode chemistries on the same model.

What Aether computes
  • P2D electrochemistry · SEI evolution · ageing
  • Cell-to-pack thermal · coolant routing
  • Thermal-runaway propagation · cell-to-cell
  • Fast-charge schedule optimisation
  • Crash · cell intrusion · short-circuit
  • Solid-state · silicon-anode · sodium chemistries
08 · Spacecraft

Satellites and on-orbit assets.

±0.04°
pointing knowledge

Thermal radiation under sun and Earth IR. Structural under launch loads and on-orbit micrometeoroid. Attitude control with reaction-wheel saturation. Solar-array deployment dynamics. Comms link-budget. The whole bus, sized and certified, in one workspace.

What Aether computes
  • Thermal radiation · sun + albedo + Earth IR
  • Structural under launch · acoustic + sine + random
  • Micrometeoroid + orbital debris (MMOD) damage
  • Attitude control · reaction-wheel saturation
  • Solar-array deployment dynamics
  • RF link-budget · ground-station handover
Coverage

Every discipline. One project file.

Each discipline ships with specialised agents, validated benchmarks and clean interop with the others — so thermal really does feed structural really does feed fatigue without a single CSV in between.

EVERY MODALITY THE WORK PRODUCESONE SIMULATION-NATIVE FOUNDATION MODEL · SHARED WEIGHTSDISCIPLINE HEADS · OUTPUTS△CAD geometry◇Mesh▦RTL · netlist⌬Molecule∿Spectra◉Plate data</>Code≋Sensorembed · every modalityself-attn · multi-headcross-modal fusionphysics-aware mlpself-attn · multi-headdecoder · discipline routingAether · shared backbone△EngineeringCFD · FEA · controls⌬Discoverymolecules · ADMET▦SemiconductorsRTL → GDS◉Autonomous labsprotocols · analysis</>Softwarewhole-repo edits⌑MaterialsDFT · band structureOne model · shared weights · trained on simulation, instrument and design data — not scraped textv1.0 · sparse-MoE · every modality · every discipline
01

Computational fluid dynamics

RANS, LES, DES, hybrid models, free-surface, multiphase, combustion, reacting flows, turbomachinery, rotor-stator, conjugate heat transfer. Validated on impinging-jet, NACA, DLR-F6 and propeller benchmarks.

02

Structural & FEA

Linear static, modal, harmonic, transient, geometric and material nonlinearity, contact, plasticity, hyperelasticity, large-deformation. Incumbent structural FEA replacement under one project file.

03

Electromagnetics

Full-wave (FEM, FEBI, FETI), MoM, FDTD, quasi-static, antenna design, SI/PI, EMC, radar cross-section. The accuracy of the leading commercial full-wave tools without their licensing.

04

Multiphysics

Coupled fluid-structure (FSI), thermal-electric, electrochemical, magneto-hydrodynamics, piezoelectric. Single mesh, single solve, no inter-tool import dance.

05

Acoustics & vibration

Modal, random, harmonic, NVH for automotive and aerospace, vibroacoustic coupling, room and underwater acoustics, sound power and transmission loss.

06

Thermal & cooling

Conjugate heat transfer, impinging jets, electronics cooling, immersion cooling, additive thermal histories, heat-pipe and two-phase flow.

07

Fatigue, fracture & crash

High- and low-cycle fatigue, strain-life, crack growth, J-integral, XFEM, explicit-dynamics crash, blast and impact at the accuracy of the leading commercial explicit-dynamics codes.

08

Additive & manufacturing

Process-structure-property models for SLM, DED, FDM, residual stress and distortion prediction, support-structure planning, calibrated to your machine.

09

Optimisation & DOE

Topology, shape, parametric, multi-objective, gradient-free, surrogate-assisted, robust-design and uncertainty quantification.

What it replaces

In place of a stack of renewals.

The list below is the typical mid-sized engineering office's CAE stack today. Aether for engineering replaces the spine. Niches — exotic foundry rules, a specific custom-cell flow — stay where they are.

Commercial structural FEA
Commercial CFD (RANS + LES)
Full-wave electromagnetics
Explicit-dynamics crash
Composite-material FEA
Free-surface + multiphase CFD
Multiphysics suites
Blast and impact codes
Industrial mesh + pre/post
Open-source CFD scripts
How it feels

A studio, not a shell script.

Aether treats simulation like a creative discipline — branching, versioned, collaborative — instead of a batch job that emails you a folder of VTU files at three in the morning.

Canvas

Geometry, mesh, physics, results — one project

Drop in CAD. Sketch the loads. Aether picks the right solver, meshes it, runs it, and explains the answer. Branch the project to try a different fillet without losing your place.

Agents

300 specialists, one conductor

A meshing agent, a turbulence agent, an EM agent, a fatigue agent — coordinated by a planner that knows when to call which. The agents are real production code, not orchestration JSON.

Validation

Benchmarks you can rerun

Every discipline ships with a regression suite of published experimental cases. Numbers move, you see why. Setup scripts and result-extraction code are part of the platform.

Solvers

Open-source where appropriate, custom where it matters

We wrap the major established open-source solvers (fluid dynamics, FEM, structural, explicit-dynamics, electromagnetics) for established workloads — and we ship our own GPU-native solvers where the commercial stack is the bottleneck.

Collaboration

Real-time, with audit

Multi-engineer co-editing on the project canvas. Comments, reviews, branching, audit trails — CAE without the lock-file dance.

Sovereignty

VPC, on-prem, air-gapped

Run in our SOC 2 Type II environment, in your VPC with private networking, or in a sovereign air-gapped configuration. The model and the geometry never leave the boundary you set.

Workflow

Six steps from CAD to signed-off result.

The path most engineers take through Aether, in the order they take it. Every step is observable, branchable, and exportable.

  1. 01

    Import

    CAD via the standard ISO neutral exchange formats, vendor-neutral kernel formats, or directly from the major commercial CAD environments.

  2. 02

    Define

    Loads, BCs, materials, contact, joints — natural-language or templated. Aether resolves to a solvable case.

  3. 03

    Mesh

    Tetrahedral, hexahedral, polyhedral. Adaptive, anisotropic, boundary-layer aware. Edits non-destructive.

  4. 04

    Solve

    Steady, transient, eigen, coupled. Streamed convergence, GPU acceleration, auto-checkpointed.

  5. 05

    Review

    Field plots, line and surface integrals, frequency response, NVH, fatigue contours — in the same canvas.

  6. 06

    Iterate

    Branch, parameterise, optimise. DOE and surrogate models built in. Everything versioned.

AI scientist

From one prompt to a finished study.

Aether is autonomous. You write the brief in one sentence; it plans the study, picks the right solvers, runs the agents, validates each artefact against measured reality, and hands you back a memo signed by the model version that produced it. We are building AI scientists, not chat assistants.

Engineer prompt

“Reduce drag on this wing by 8% while keeping structural margin above 1.5 and respecting our manufacturing constraints. Hand me a CAE memo at the end.”

— staff aerodynamicist · 09:14
  1. 01Prompt

    One sentence from an engineer

    “Reduce drag on this wing by 8% while keeping structural margin above 1.5 and respecting our manufacturing constraints. Hand me a CAE memo at the end.”

  2. 02Plan

    Aether proposes the study

    The planner agent decomposes the brief into a study tree — geometry sweep, CFD, FEA, manufacturability check, optimiser loop — picks the right solvers and posts the plan for your approval before any compute runs.

  3. 03Execute

    Specialist agents run the work

    Mesh agents prepare the discretisation. CFD agents resolve the boundary layer and run the polar. FEA agents solve the structural loadcase. Optimiser agents drive the surrogate. Each is a real production module — not orchestration glue.

  4. 04Validate

    Every artefact against a benchmark

    Before any result reaches you, it's compared against the regression suite for that discipline: a known benchmark case, a measured wind-tunnel point, or a previous customer-validated study. If the comparison drifts, the run halts and surfaces a citation.

  5. 05Report

    A signed CAE memo with the receipts

    Pressure contours, mode shapes, stress fields, optimisation history, the manufacturability margin table, the convergence log and the model-version hash that produced each. Ready for design review without rebuilding the deck.

Time end-to-end
4 h 12 m
Wall-clock from the prompt above to a signed memo, on a single H100 node.
Agents that ran
17
Mesh, CFD, FEA, optimiser, manufacturability check, reviewer, scribe — each a real production module.
Drag reduction
8.4%
Achieved while margin held at 1.62 and every geometry feature stayed inside your machining envelope.

The same shape runs for a chip RTL block, an antibody library, a wet-lab active-learning campaign, a tokamak operating scenario or a launcher trajectory. Aether is the model the discipline-specific agents share. The disciplines vary; the autonomy is the constant.

Why the cycle shortens

Months of R&D, compressed to hours.

The slowest part of engineering R&D isn't the simulation — it's the hand-offs around it. Setting up the case. Waiting for the queue. Re-meshing. Writing the memo. Aether collapses each of those into the autonomous agent flow. Same answer, the right side of a quarter.

Stage
Today
With Aether
Why
Set up the case
5–9 days
minutes
Geometry import, mesh, BCs, materials and solver settings happen as the model reads your brief — not as a week of ICEM and journal scripts.
Run the study
1–4 weeks
hours
GPU-native solvers + the right solver picked by the planner. The DOE loops in the background; you read a finished trade-study instead of running each point.
Validate and review
1–2 weeks
same day
Every artefact compared against a benchmark before it reaches your inbox. Failures surface a citation, not a 6 a.m. email.
Write the memo
3–5 days
minutes
The scribe agent writes the CAE memo as the run completes — with figures, tables, margins, convergence and the model-version hash.
Time to first result
60–100×

vs. a senior engineer driving the incumbent CAE stack by hand on the same loadcase. The compression compounds with study count.

Engineer time saved
~70%

Median across the first wave of customer pilots. Reclaimed time goes to the parts of the design only humans do well — judgement, trade-offs, taste.

Product to market
−9 months

On a typical mid-volume hardware programme. The dominant gain is in design iteration count per quarter — not in any one solve being faster.

AI scientists, not chatbots

Aether is an autonomous scientist.

We're not wrapping a chat model around your existing CAE. We're training a foundation model that does the work — plans the study, picks the solver, runs the agents, validates the artefact, writes the memo. Six things every senior engineer used to do by hand. Now Aether does them.

  • 01

    Plans the study

    Decomposes a one-sentence engineering brief into a directed acyclic graph of agent calls — meshing, solving, optimising, validating, reporting — and posts the plan for your approval before any compute runs.

  • 02

    Picks the right solver

    Knows when RANS is enough and when DDES is required. Knows when implicit beats explicit. Knows which turbulence closure pairs with which boundary layer. The judgement calls a senior engineer makes — encoded in the model.

  • 03

    Runs the work

    Specialist agents execute the plan in parallel. Each is a real production module with stable contracts — not orchestration JSON. Long-horizon runs check-point every step.

  • 04

    Validates against reality

    Every artefact compared against the regression suite for that discipline: a published experimental case, a customer-validated study, or your own historical wind-tunnel data. Drift halts the run.

  • 05

    Writes the report

    A signed CAE memo with figures, tables, margins, optimisation history, convergence and the model-version hash that produced each. Replayable by anyone with the run ID.

  • 06

    Improves itself

    Failed cases enter the eval corpus. Successful pilots become regression tests. The model that solves your next study is better than the one that solved this one — and you have the diff.

Where this leads

A foundation model that does the work, not just the chat about it — the substrate for general physical intelligence.

Aether is built to think in physics, chemistry, biology and silicon — the substrate of the physical world. We are moving toward general intelligence through the disciplines that have to face reality. The model that designs a rocket today is the model that runs a lab tomorrow.

The outputs you actually get

Fields, plots, tables — signed and exportable.

The post-processing artefacts an engineer expects from a commercial CAE suite, generated in the same project canvas the run happened in. Every figure carries the model version that produced it.

  • Pressure contours

    Cp, p, total-pressure-loss fields on surfaces and cut planes. Annotated, with colorbars and reference state.

  • Streamlines & vector fields

    Steady and instantaneous streamlines, line-integral convolution, particle traces, vortex cores extracted with Q-criterion.

  • Mode shapes

    Eigenmodes with natural frequencies, animated deformation, modal mass and effective-mass tables — exportable to your NVH report.

  • Stress fields

    Von Mises, principal, shear and Hill stress contours. Peak locations called out. Margin-of-safety tables against your material allowables.

  • Convergence histories

    Per-equation residuals on log scale, monitor histories, tolerance lines. The run halts when the convergence criterion holds, not when the wall clock runs out.

  • Frequency response

    Magnitude and phase of the transfer function from input to monitored output. NVH, control, antenna, vibration — same shape, different units.

  • Probe time histories

    Pressure, strain, temperature, displacement and force probes sampled at user-defined points. CSV/Parquet export, Pandas-ready.

  • Electromagnetic fields

    |E|, |H|, current density, S-parameters and far-field radiation patterns. FDTD, FEM and MoM — extracted in the same canvas.

  • Meshes

    Unstructured tets, polyhedra, hex-dominant. Boundary-layer prisms with first-cell y+ control. Mesh quality reports as a first-class artefact.

Export formats

VTK, VTU, CGNS, EXODUS, HDF5, Parquet, CSV, PDF, PowerPoint and the major commercial CAE formats your design-review audience expects.

Reports

Auto-written CAE memos with figures, tables, margins-of-safety, the convergence log and the model-version hash — ready for design review.

Audit

Every artefact is signed by the model version, the agent ID and the capability grants that produced it. Replayable by hash.

Validation

Numbers against measured reality.

We benchmark against measurements, not against other models. Each row below is reproducible from the platform — the case, the mesh, the solver settings, the extraction scripts.

Cooper impinging-jet (Nu)
3.6%
mean error
NACA-0012 transonic drag
1.8%
vs experiment
Frigus heat-pipe
2.4%
wall ΔT error
DLR-F6 cruise lift/drag
≤1.0%
polar deviation
Notched bar fatigue (ASTM)
0.92
Pearson on life
Helmholtz cavity acoustics
1.1%
modal frequency error
Interop

Reads what your team already exports.

We don't expect anyone to throw away ten years of CAD, mesh and post-processing tooling on day one. Aether reads the formats your team already uses.

ISO neutral CAD exchange
Vendor-neutral kernel formats
Major commercial CAD
Commercial structural input decks
Open-source CFD case dictionaries
VTK / VTU / CGNS
MED / open structural formats
Open explicit-dynamics input
Open FEM exchange formats
EXODUS
Commercial structural script formats
Commercial CFD journal scripts
Vs the CAE stack

Every solver, one model.

The commercial CAE suite is one vendor per physics and an import in between. Aether solves them together on one mesh — validated against measured reality, not just each other.

✓Structural FEA
✓CFD
✓Electromagnetics
✓Multiphysics, one mesh
✓Explicit dynamics / crash
✓Natural-language setup
✓Calibrated uncertainty
✓Validated against measured reality
✓One model, every discipline
Related research

The work behind this product.

We publish what we learn. The posts below are the substantive notes behind this product — methods, evaluations, case studies.

Methods

Cross-domain transfer: from turbulence to transcriptomics

Read
Engineering

How Aether retired a customer's seven-figure CAE renewal

Read
Benchmark

Beating commercial CFD on impinging-jet validation

Read

What would you stop paying for first?

Tell us which CAE renewal is the worst line on your budget. We'll port one project across and let the numbers speak.

Start a pilotUnder the hood: Aether