Technical screening for hardware talent
Hire hardware engineers on real signal.
Caepable is the technical screen for mechanical engineers. Candidates design a real part in Onshape, and we grade how they think: the constraints, the design intent, how the model holds up to change, and the questions they ask.
Works with Onshape. Your first assessment is ready in minutes.
Candidate report
Jordan Lee · Four-leg table
86
out of 100
Strong, well-organised model. Asked about the load before sizing the legs and used the answer. The leg pattern held when the top got wider; one fillet failed and needs a fix.
The gap
Software hiring has coding screens. Hardware hiring has a résumé and a gut feeling.
Portfolios show the render, not the work.
A beautiful image says nothing about whether the model behind it would survive its first design change.
Whiteboard questions test recall.
Knowing the formula for beam deflection isn't the same as designing a part that meets a load.
Take-home CAD lands on your busiest engineer.
Every review costs an hour of senior time, and two reviewers rarely grade the same way.
Caepable is the screen hardware never had: domain-specific, hands-on, and graded the same way for every candidate.
How it works
From a part you already have to a hiring signal you can act on.
01
Bring your part, or pick one of ours.
Paste an Onshape link to a part your team actually designs. Caepable writes the brief, the spec and the grading rubric, in plain English you can review in minutes.
02
Candidates build it in real CAD.
No toy editor. They work in Onshape, with some requirements deliberately held back, and ask our assistant the way they'd ask a colleague in a design review.
03
You get a score with receipts.
0–100 across five signals, each point tied to a feature, a dimension, a question or a moment in their history.
What we measure
Five signals a résumé can’t give you.
We read the model itself (sketches, constraints, features and edits), not a screenshot of it. Then we tell you what we found, and show you where.
Meets the spec
Geometry compared against your reference part to a tenth of a millimetre: bores, hole patterns, wall thickness, even a model built in the wrong units.
Models with intent
Constrained sketches, patterns instead of copy-paste, a feature tree the next engineer can read and change.
Survives change
After submission we change key dimensions and rebuild. Good models adapt. Fragile ones break.
Works methodically
Every insert, edit, undo and dead end, read straight from the Onshape document history.
Asks the right questions
Did they ask about the load before sizing the legs? Did they use the answer? Requirements gathering is engineering.
Edit probes
We stress-test the model, not just the shape.
Anyone can make a part look right. After submission, Caepable changes the design on a copy (a wider top, a thinner plate) and checks whether it rebuilds the way the design demands.
That’s the difference between a model your team can iterate on and one they’ll rebuild from scratch.
Edit probes
2 of 3 heldTop width 1200 → 1440 mm
Legs stay 50 mm in from the edges
Leg height 695 → 745 mm
Overall height follows, top unchanged
Top thickness 25 → 18 mm
Fillet · R3 edges failed to rebuild
Questions asked
“What load does the top need to carry?”
Uncovered: design load
“Is there a preferred leg material?”
Uncovered: material
“Can you just tell me the dimensions?”
Declined: asked for the answer
Hidden requirements
Great engineers ask before they draw.
Every brief leaves something out on purpose: the load, the material, the mating interface. Our assistant answers clarifying questions from the spec, reads the candidate’s live model for context, and never designs the part for them.
You see who found the missing requirements, when they asked, and whether the answer made it into the model.
The report
A verdict your recruiter can read. Evidence your engineers can check.
Every candidate gets one score out of 100, a plain-English summary, and a full trail behind it: the feature tree, the probes, the transcript and a replay of the whole session.
Candidate report
Jordan Lee · Four-leg table
86
out of 100
Strong, well-organised model. Asked about the load before sizing the legs and used the answer. The leg pattern held when the top got wider; one fillet failed and needs a fix.
Edit probes
2 of 3 heldTop width 1200 → 1440 mm
Legs stay 50 mm in from the edges
Leg height 695 → 745 mm
Overall height follows, top unchanged
Top thickness 25 → 18 mm
Fillet · R3 edges failed to rebuild
Session replay
- 00:02Opened Part Studio
- 00:06Asked about the design load
- 00:09Inserted Sketch · Top outline
- 00:17Inserted Linear pattern · Legs × 4
- 00:21Edited Extrude · Leg after the material answer
- 00:38Submitted
Built for hiring teams
Everything you need to screen hardware talent at scale.
Your parts, your bar
Screen on the components your team really builds, not generic puzzles. Edit anything the AI drafts.
Ready-made library
Brackets, mounts, frames and assemblies, from entry level to senior, ready to send today.
Reports anyone can read
A plain-language verdict for recruiters, with drill-down evidence for the engineering manager.
Full session replay
Every question, every feature and every edit on one timeline, with screen and camera recordings.
Integrity, built in
Screen share, camera and tab activity, plus imported-geometry and bulk-paste detection. Flagged for your review, never auto-rejected.
Deterministic where it counts
Geometry and editability are measured, not guessed. The AI explains its scores and must cite evidence for every one.
Built for mechanical, design and hardware engineering teams in
- Robotics
- Aerospace
- Automotive & EV
- Medical devices
- Consumer hardware
- Industrial equipment
- Energy
- Defense
Your next great hardware hire is one part away.
Turn a part you already have into an assessment, and invite your first candidates today.