Dataset catalogue

Engineering data, from the firms that made it.

Native CAD, CAM, CAE and scan data from real engineering work, de-identified and licensed for AI training. Every dataset below has a sample we can send. Tell us what you are building and we will share the right one.

8Datasets

6Samples ready

6Domains

Automotive · Body-in-white Available

SUV body programme

Automotive design programme

Native CATIA V5 and NX data from a complete SUV body programme: BIW structure, closures, pillars and the full-vehicle model. Parts ship with their assemblies, so the structure of a real vehicle stays intact.

Volume: ~1,260 native parts (≈500 CATIA V5, ≈760 NX), 121 assemblies, full-vehicle model · 21.5 GB
Sample: 5- or 10-part pack: native CATIA, STEP and 6 renders per part
Timeline: Full set within 2 weeks of agreement

RL tasks this supports

Rebuild a part feature-by-feature to match a target STEP
Assemble BIW sub-structures from parts
Change a section and update the mating parts
Surface repair and clean-up

CATIA V5 · NX · STEP · IGES · STL Request sample

Mechanical · Parts, assemblies, drawings Available

Mechanical design archive

Engineering design programmes

Parts, assemblies and 2D drawings from structured design programmes: mechanisms, valves, fixtures and engine sub-systems. Each assembly ships with every part and its drawings.

Volume: ~18,400 unique parts (13,862 CATIA, 3,131 NX/Creo, 1,402 SolidWorks) · 2,781 assemblies
Sample: 10 complete SolidWorks assemblies with parts and drawings
Timeline: 2.5 GB organised set available now; full archive in 3–4 weeks

RL tasks this supports

Model a part from its 2D drawing
Assemble from parts and constraints
Create drawings with GD&T
Set up and check mechanism motion

CATIA V5 · SolidWorks · NX · Creo · STEP Request sample

Reverse engineering · Industrial parts Available

Industrial reverse engineering: scan to CAD

Reverse-engineering services firm

Matched sets for real industrial parts: raw point cloud, scan mesh, the Design X rebuild, native SolidWorks and STEP. Each part is graded easy, medium or complex.

Volume: Sample: 7 parts, 1.9 GB (10 planned) · the firm has delivered ~1,500 projects, scope to follow
Sample: Per part: point cloud, mesh, Design X, SolidWorks and STEP
Timeline: Sample parts available now; further parts scoped on request

RL tasks this supports

Point cloud to parametric CAD
Recognise features in a scan mesh
Transfer a Design X rebuild to SolidWorks
Grade CAD against the scan for accuracy

Point cloud (.pts) · STL · Design X (.xrl) · SolidWorks · STEP Request sample

Software · Code Available

Execution-verified code at scale

Software company

About 2,050 programming problems with specifications, reference solutions, curated test suites and a working execution environment, plus 467 million real solution attempts that can be replayed step by step up to a passing solution.

467MSolution attempts

+5MPer month

2,050Problems

35+Languages

Volume: ~2,050 problems · 467M solution attempts, growing ~5M a month · Python 43%, Java 24%, SQL 17%, C 9%, C++ 8%, 30+ more
Sample: 75 problems with 852 test cases and 750 passing and failing solutions
Timeline: Available now; tranches by language, topic or difficulty

RL tasks this supports

Write code that passes a hidden test suite
Learn from attempt sequences that end in a passing solution
Debug a failing solution
Agent traces (about 3% of problems)

Specifications · Test suites · Source code · Execution results · Agent traces Request sample

Automotive · Reverse engineering Sample ready

Sedan interior trim: scan to CAD

Automotive reverse-engineering project

Interior door-trim parts rebuilt from 3D scans into production CAD. Two parts include the original scan mesh alongside the finished model.

Volume: 5 parts · 2 scan meshes (6.2M and 9.3M triangles)
Sample: All 5 parts: STEP, Parasolid, CATIA where available, 10 renders each
Timeline: Available now

RL tasks this supports

Reconstruct freeform surfaces from a scan
Fit and trim class-A surfaces
Check CAD deviation against the scan

STEP · Parasolid · CATIA V5 · IGES · STL Request sample

Civil · Architecture, structure, MEP Sample ready

Hospital building: drawings to BIM

Architecture practice

A complete G+6 hospital block: architectural, structural, MEP and finishes drawings. Made for 2D-to-3D BIM tasks, with the Revit model requested as ground truth.

Volume: 13 DWG + 3 PDF architectural and MEP sheets · 9-sheet structural construction set
Sample: One built task: drawings and brief to IFC model
Timeline: Drawings available now; Revit ground truth pending

RL tasks this supports

Build a Revit / IFC model from drawings
Lay out structure from construction drawings
Coordinate MEP against architecture

AutoCAD DWG · PDF · IFC Request sample

Manufacturing · CAM Coming soon

CNC machining programmes

Precision machining shops

Complete Mastercam and NX CAM jobs: the finished part, stock and workholding, every operation and tool, and the posted G-code the machine actually ran.

Volume: First jobs in review
Sample: Per job: native CAM file, posted NC programs, tool sheet, STEP of the finished part
Timeline: First packaged jobs expected within 4 weeks

RL tasks this supports

Choose a machining strategy for a part
Select tools and holders
Verify G-code in simulation
Fix a gouge or collision

Mastercam · NX CAM · G-code · STEP Request sample

Automotive · CAD, CAE, manufacturing Fully licensed

Automotive engineering services archive

Automotive engineering services firm

Fifteen years of production engineering work: sheet-metal and plastic part design, fixtures and automation, CAE models, and the project email that ties each request to what was delivered.

Volume: ~1.9 TB unique CAD and CAE · ~136,000 native parts · 164,000 engineering emails since 2011
Sample: SolidWorks automation units; CATIA sheet-metal and plastic door panels
Timeline: Fully licensed. We are sourcing similar engineering-services archives; ask to be first in line

RL tasks this supports

Design a part from a real change request
Sheet-metal and plastic part design
CAE pre-processing (HyperMesh, LS-DYNA)
Long-horizon design-revision trajectories

CATIA V5 · SolidWorks · NX · STEP · JT · HyperMesh · LS-DYNA Ask about similar data

Request a sample

Tell us what you are building

We reply within two working days. Samples are shared directly once we understand the use case, and every licence is for AI training only.