Your customer audits the
process, not just the part.
AI inspection for automotive production — surface finish, assembly completeness and component verification, with a record for every part that leaves the cell.
Test it on your productThe Hard Case
A missed defect is not one part.
It is every part since the last check.
Automotive quality is containment as much as detection. When something escapes, the question is not what went wrong with that part — it is which parts shipped between the last good check and this one, and whether you can prove it.
Sampling cannot answer that. A per-unit record can, which is why the value of inspecting every part is often less about catching more defects and more about bounding the problem when one gets through.
The inspection itself is also harder than it looks. Machined and coated automotive surfaces are specular, geometry varies across a family of parts, and the difference between an acceptable tool mark and a reject is frequently a judgement your best inspector makes and cannot fully write down.
Containment is bounded by your record.
A per-unit result with an image reference turns an escape from an open-ended recall question into a defined set of serials or a defined window of production.
Part families vary; the check should not.
One model handles several part variants on the same line, so a cell running a family does not need reconfiguring between them.
What it checks
The checks an automotive cell
actually needs.
Is the finish acceptable?
Surface defects
Scratches, dents, porosity, tool marks and coating flaws on machined, cast and painted surfaces.
Is everything fitted?
Assembly completeness
Presence, position and orientation of clips, fasteners, seals, connectors and sub-components.
Is it within tolerance?
Dimensional check
Key dimensions and geometry recorded as values for every part, so drift is visible before parts go out of tolerance.
Is the joint sound to look at?
Weld & joint
Visual weld appearance, spatter, seam position and joint condition on visible welds.
Is this the right part?
Component identification
Confirms the variant running matches the job, catching mixed parts before they reach assembly.
Is the part marked and readable?
Marking & DMC
Part marking, data matrix and label reading verified against expected content for traceability.
Is the bead right?
Sealing & adhesive
Presence, continuity, width and position of adhesive and sealant beads.
Is anything on or in the part?
Foreign material
Swarf, debris and contamination on machined faces and in cavities before assembly.
Is it presented correctly?
Orientation & position
Part orientation and placement in fixtures, trays and dunnage before the next operation.
Built for the environment
Oil, swarf, vibration and a
line that does not stop.
Automotive cells are mechanically hostile and commercially unforgiving on downtime. These are the conditions the hardware is specified against.
ENVIRONMENT
Built for stability. Designed for real-world conditions.
- ContaminationSealed housings against oil mist, coolant and swarf
- VibrationMounting isolated from press, robot and conveyor movement
- HeatPlacement and shielding for weld and forming operations
- Cycle timeInspection specified to fit within the cell cycle, not extend it
- GuardingMounting designed around existing guarding, robots and access
IMAGING
Clearer vision. Smarter inspection.
- Specular surfacesControlled lighting geometry for machined and coated faces
- Geometry and depth3D sensing where form and height matter beyond the outline
- Fine surface defectsResolution matched to the smallest defect at working distance
- Marking and DMCOCR, OCV and code reading on marked parts
- Multiple facesSeveral inspection points or reorientation for all-round coverage
- Dark and coated partsIllumination selected for low-contrast surfaces
On lighting. Machined and coated automotive surfaces are specular, which means a fixed light produces a different image on every part presentation. Lighting geometry is specified per part family and validated against real parts, including the awkward ones, during the trial.
Where the record matters
Traceability is what your
customer is really buying.
Automotive quality systems are built around demonstrable process control and the ability to contain a problem quickly. Inspection records are the raw material for both.
Part-level traceability
Every part inspected is logged with its result, its measured values and an image reference, so a question about a specific serial or window has an answer.
Process evidence
Automotive quality management expects demonstrable control rather than assertion. Continuous per-unit data is stronger evidence than periodic sampling records.
Containment speed
When an escape happens, the record defines the affected window. Without it, containment expands to whatever cannot be ruled out.
The Platform
One Place to Train.
Two Place to Run
Shared by both
Model development
Models
Quality
Line 1
Line 2
Line 3
Results
Every line, plant and country
Safety
Production floor
existing cameras
Warehouse
existing cameras
Loading bay
existing cameras
Streams
Runtime and dashboard in one
Standards & codes
The standards your parts
are judged against.
The published criteria your customers, auditors and quality team already reference. We train models against your interpretation of these standards — we do not certify conformance to them, and an inspection result is not a conformity assessment.
ISO 5817
What it covers
Welding — Fusion-welded joints in steel, nickel, titanium and their alloys — Quality levels for imperfections, at Levels B, C and D.
Where it applies
Visual weld appearance on visible joints
ISO/IEC 15415 & TR 29158
What it covers
2D code print quality and the direct part mark variant, graded A to F, with additional lighting angles for marked metal surfaces.
Where it applies
Data matrix verification for part traceability
ISO/IEC TR 29158
What it covers
Classification of geometric imperfections in metallic materials for fusion welding, providing the designation system used to name each imperfection.
Where it applies
Naming the weld defect your drawing refers to
Automotive questions
What automotive quality team ask first.
Yes. A single xis.rt instance handles several variants at once, each with its own checks, so a cell running a part family does not need reconfiguring between them. New variants are added by your own team in xis.dev.
Still deciding
The one question this page can't answer:
does it work on your defect?
Start here
Send us your images
Share examples of the defect you care about. We train a model on your product and come back with a straight answer — including if it's no.
Start a feasibility checkOr ask directly
Talk to an engineer
Thirty minutes with someone who builds this. Bring your line speed, your defects and your integration constraints.