From Raw Data To State-of-the-art Results On Aerodynamics ML With Neural Concept

• The AI-first and enterprise-ready engineering platform has demonstrated state-of-the-art accuracy on the recent DrivAerNet++ dataset.
• The project was executed in less than a week from start to finish using standard components available in the Neural Concept platform.
• The key was using the Neural Concept platform to transform 39 terabytes of aerodynamic simulation data, including 8,000 car geometries, engage into a large-scale automated data preprocessing and train large deep neural network models.
• Neural Concept customers include leading automotive OEMs, who have achieved up to 30% shorter design cycles and $20 million in savings on 100,000-unit vehicle programs.

Lausanne, Switzerland – September 11, 2025 - Neural Concept, the leading AI-first engineering platform for product development, today published a new study on the industry-recognised DrivAerNet++ dataset for automotive aerodynamics. The work sets new standards for Aerodynamics predictions using deep-learning, demonstrating the platform’s ability to process large-scale industry datasets, extract geometric encodings, train models and deliver state-of-the-art results, all using its default settings. This milestone reflects years of technology development in partnership with leading automotive OEMs.

Designing vehicles that are both energy efficient and cost-effective has become a critical challenge for the automotive industry. Even small reductions in aerodynamic drag can translate into significant fuel savings or extended range for electric vehicles. But achieving these gains has traditionally required long, expensive design cycles built around high-performance computing and complex simulations.

‍Setting a New Standard in AI-Driven Aerodynamics

Using this benchmark academic dataset for automotive aerodynamics testing, Neural Concept transformed 39 terabytes (TB) of simulation data — comprising over 8,000 car geometries — into a full end-to-end workflow, from model training to real-time deployment within a week, a process which would typically take months to complete using other tools. Using its default model configuration, the platform outperformed previously published methods in prediction accuracy surface pressure, wall shear stress, volumetric velocity, and drag coefficient, without requiring any parameter tuning.

Neural Concept continues to cement its platform as the leader in AI for automotive design and development, not only by achieving unmatched model accuracy, but by turning models into reliable, auditable and evolvable enterprise-grade systems, seamlessly deployed into the tools where engineers work.

"We built DrivAerNet++ as an academic foundation to accelerate automotive transformation in the AI era,” said Dr. Faez Ahmed, Associate Professor, Department of Mechanical Engineering, Massachusetts Institute of Technology (MIT). “It’s thrilling to see it quickly picked up by companies like Neural Concept to redefine the speed of real industrial design cycles."

‍Measurable Industry Impact

Neural Concept’s platform is delivering measurable industrial impact, with customers such as Subaru, Bosch, and OPmobility, as well as major aerospace, defense, consumer electronics, and semiconductor players, reporting up to 30% shorter design cycles and $20 million in savings on 100,000-unit vehicle programs.  

“Delivering top performance on a widely recognised benchmark using the default model configuration reinforces the robustness and maturity of our platform,” said Théophile Allard, Chief Technology Officer at Neural Concept. “But what was most satisfactory is to see how engineering teams can turn gigantic, complex datasets into production-ready workflow reliably.  This is the meaning of an enterprise Engineering Intelligence platform: turning state-of-the-art AI into reliable and collaborative tools which accelerate vehicle programs and deliver valuable cost savings.”

‍From Prediction to Production

Beyond model innovation, real engineering value comes from deployment, turning AI models into collaborative, auditable workflows that evolve across teams and projects. Neural Concept’s ‘Design Lab’ does just that: embedding AI design copilots directly into engineering tools, where they provide real-time performance feedback, geometric improvement suggestions, and always up-to-date KPIs to keep pace with fast-moving vehicle programs.

Integrated with NVIDIA Omniverse for live design visualization and deployed on Microsoft Azure for enterprise-grade scalability, the Neural Concept ‘Design Lab’ transforms AI models into interactive co-pilot experiences that process industrial datasets in minutes, eliminating capacity planning and putting Neural Concept’s vision of Engineering Intelligence into practice.

“Neural Concept’s breakthrough demonstrates the power of combining advanced AI with the scalability of Microsoft Azure,” said Jack Kabat, Partner, Azure HPC and AI Infrastructure Products, Microsoft. “By running training and deployment on Azure’s high-performance infrastructure — specifically the NC H100 Virtual Machine— Neural Concept was able to transform 39 terabytes of data into a production-ready workflow in just two weeks. This shows how Azure accelerates innovation and helps automotive manufacturers bring better products to market faster.”

Neural Concept has announced that this breakthrough workflow is now available to all customers as part of the Neural Concept Community Kit, providing an accessible starting point for industrial-scale AI deployment in automotive and other sectors.  

Learn more about how we achieved the results in our blog article here.

About Neural Concept

Founded in 2018, Neural Concept provides the leading AI-first engineering platform for product development. By embedding AI natively into design and simulation workflows, Neural Concept empowers engineering teams to compress development cycles from months to days, improve product performance across efficiency, safety, and sustainability, and scale AI adoption without costly, years-long integration.  

The company partners with more than 70% of the world’s largest OEMs and 40% of the top 100 tier-1 suppliers across automotive, aerospace, defense, consumer electronics, and semiconductors. Neural Concept was spun out of the Swiss Federal Institute of Technology in Lausanne (EPFL) and is backed by global investors, including Forestay Capital and D. E. Shaw Ventures.

For more information, visit www.neuralconcept.com.