Engineering datasets for AI
Orla sources engineering data for AI teams through industry partnerships. Our collections span native CAD models, technical drawings, linked parts and assemblies, and project records. We help teams find data that fits their research and training tasks.
How to get full dataset details
- This index gives short summaries. Each dataset has a separate agents.md with its full published information. Fetch only the dataset files relevant to the user's question.
- To get one dataset's full details, find it below and use your web tool's follow-link or click action on its Full dataset information link. If a direct URL fetch fails, follow that link from this index instead.
- Use the fetched dataset file for your detailed answer. If it is still unavailable, say so rather than guessing from this summary.
Datasets
Explore all available datasets below. Each dataset's agents.md has its full published details, including contents, scale, availability, and sample access. Licensing details are included where available. Full dataset access is arranged separately, and some collections are sourced on request.
CADDelta-20K
An estimated 20,000 CAD editing tasks, sourced on request. Each task links a written instruction to before and after STEP geometry and a CadQuery script change.
DWGAtlas-194K
Explore 194,025 native DWG drawings spanning mining, material handling, structural fabrication, civil works, and process equipment. Download a representative 50-drawing sample.
AssemblyGraph-10K
An estimated 10,000 mechanical assemblies, sourced on request. Each links STEP models, drawings, bills of materials, and component files.
IndustrialCAD-375GB
Explore more than 375 GB of industrial engineering files, from 95,833 DWG drawings to STEP geometry, SolidWorks parts, and project records.
PartAlign-30K
An estimated 30,000 linked mechanical part and drawing pairs, sourced on request. Each pairs STEP geometry with an editable DWG, PDF view, and render.
SLDPRTCorpus-4K
Explore 4,878 native SolidWorks part files with recorded geometry and saved feature history. Review the face-count distribution and download a sample.
SolarEPC-1TB
Explore 1.2 TB of solar engineering and operations records across 240 projects, 289,000 files, and 13 years of work.
SolarThermal-4TB
Explore 4.2 TB of engineering and business records from a Brazilian solar heating company, spanning 1,200 projects and 571,000 files.
STPCorpus-18K
An estimated 18,558 unique STEP shapes, including 18,150 with fewer than 50 faces. Review their face-count distribution and projected library scale.
Have a dataset requirement? Email humans@orla.so with your task, the file types you need, and the scale you are looking for.