Vorelios
Vorelios provides a machine-learning model that predicts aerodynamic flow fields—pressure, skin friction, lift, drag, and moments—over unseen geometries in milliseconds, replacing hours-long high-fidelity CFD runs. The model is trained on internally generated high-fidelity simulation data and demonstrates lower prediction error than published models across three public benchmarks, while running 6–35× faster than Transolver-3 and AB-UPT on the same GPU. The company is extending its model beyond aerodynamics to heat transfer and structural mechanics for multi-physics design evaluation.
- Aerospace
- Artificial Intelligence
- Software Only
Funding
Raised to date
$500KRaised to date based on public sources. This may differ from the amount the company actually raised and is based only on what is publicly available on the internet.
Founders
Founder details are not available yet.
Product
Problem
Aerospace engineering teams rely on high-fidelity computational fluid dynamics (CFD) solvers to evaluate aerodynamic designs, but each simulation run takes hours. This computational bottleneck limits the number of design variants teams can explore, forcing engineers to make tradeoffs between design breadth and simulation accuracy.
Solution
Vorelios provides a machine-learning model that predicts aerodynamic flow fields—pressure, skin friction, lift, drag, and moments—over unseen geometries in milliseconds, replacing hours-long high-fidelity CFD runs. The model is trained on internally generated high-fidelity simulation data spanning a wide range of geometries and operating conditions, allowing it to generalize to designs it has never seen. Integrating the predicted surface fields yields integrated quantities like lift and drag, enabling engineers to update designs and instantly see the full physical response. The model is over 1,000,000× faster than a high-fidelity solver, evaluating all 2,500 aircraft in the BlendedNet++ test set in under 5 seconds. The company is extending its model beyond aerodynamics to heat transfer and structural mechanics, enabling multi-physics design evaluation within a single framework.
Target Audience
Aerospace engineering teams whose design workflows are constrained by CFD simulation turnaround times, including those performing design-space exploration, trade studies, and design optimization.
Features
- Predicts full surface pressure and skin-friction fields on unseen geometries with a single model call, eliminating per-design solver runs
- 1–31 ms per design on an NVIDIA H100, achieving 6–35× faster inference than Transolver-3 and AB-UPT on the same GPU
- Demonstrates lower prediction error than published models (Transolver-3, AB-UPT, GeoTransolver, CANTO) on NASA Common Research Model, BlendedNet++, and HiLiftAeroML public benchmarks
- Generalizes across diverse geometries and operating conditions, including transonic transport aircraft, blended-wing-body configurations, and high-lift designs through stall
- Roadmap extends to heat transfer and structural mechanics, enabling coupled multi-physics evaluation with the same model