Normal
Normal is an applied research lab that advances frontier AI capabilities for hardware design by creating specialized datasets, benchmarks, and technical reports. The company builds evaluation frameworks such as CAD Arena and FrontierCAD, which test native agent drawing-to-CAD performance and score frontier models on 44 parts and 16 assemblies made from engineering drawings. These tools provide quantitative comparisons that help engineers assess model capability for rockets, robots, aircraft, satellites, medical devices, and semiconductors.
- Artificial Intelligence
- Data & Analytics
- 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
Engineering teams designing physical products like rockets, robots, and medical devices increasingly rely on AI for CAD generation, yet there are no standardized benchmarks or datasets to evaluate how well frontier models perform on real hardware design tasks. Without reliable evaluation methods, engineers cannot trust or adopt AI tools for complex mechanical and electrical design workflows.
Solution
Normal is an applied research lab that advances frontier AI capabilities for hardware design by creating specialized datasets, benchmarks, and technical reports. The company builds evaluation frameworks such as CAD Arena and FrontierCAD, which test native agent drawing-to-CAD performance and score frontier models on 44 parts and 16 assemblies made from engineering drawings. These tools verify AI-generated parts and assemblies, providing quantitative comparisons that help engineers assess model capability. Normal also offers unrestricted access to its benchmarks and datasets to qualified partners, supporting the wider hardware development ecosystem.
Target Audience
Primary customers are engineers and researchers building rockets, robots, aircraft, satellites, medical devices, and semiconductors who need reliable AI tools and standardized evaluation methods for hardware design workflows.
Features
- CAD Arena: a native agent drawing-to-CAD benchmark with verified reference and output comparisons for evaluating model performance
- FrontierCAD: a verifier study that improves the part verifier, adds a verifier for assemblies, and scores frontier models on 44 parts and 16 assemblies from engineering drawings
- Publication of technical reports and datasets on machine-generated CAD for use by engineers and researchers
- Direct access to benchmark suites and datasets for partner organizations