Stable decision
Agreed thresholds and one OK/NOK/Error logic instead of subjective judgement.
On-premises AI visual inspection
Quality control that sees more.
Automate manual quality inspection without replacing your entire production environment. AIVQC combines data, training, validation and local OK/NOK decisions in one controlled process.
Reuse existing hardware where it meets the requirements. Data can stay inside your plant. Start with one inspection.
- ✓ No mandatory cloud
- ✓ Operates offline on the line
- ✓ Controlled model versions
Image captureLocal analysisDecision and trace
Interface demonstration — not a reference validation result.
DATA UNDER CONTROLProduction images can stay inside your plant
DECISION ON THE LINEThe OK/NOK result is created locally
FLEXIBLE HARDWAREWe check what you already have first
TRACEABILITYModel, recipe and configuration are recorded
More complete than a smart camera.
Less complex than enterprise machine vision.
WHY AIVQC
From manual inspection to a repeatable decision.
An operator can see a defect. AIVQC helps turn that knowledge into a measurable, versioned process for every captured part.
We do not sell “just a model”. We combine imaging, data, validation, deployment and operational line support.
Your hardware first
We do not assume the computer, camera or entire station must be replaced. We qualify the existing setup first.
Data under control
Production images do not have to leave your infrastructure.
Full traceability
You know which model, recipe and configuration produced every result.
RETROFIT, NOT REVOLUTION
We check the hardware you already have first.
AIVQC qualifies the computer, camera and configuration for a specific process instead of requiring one closed hardware ecosystem.
- 01 Camera, optics and lighting assessment
- 02 Decision-time measurement on the target computer
- 03 Replacement recommendation only where necessary
HOW THE PLATFORM IS BUILT
From the first image to the decision on the line.
Three specialised components prepare, publish and run an approved inspection.
01TRAIN
AIVQC Trainer
Prepare data and verify inspection quality before release.
- +Images and annotations
- +Training and comparison
- +Acceptance criteria
02CONTROL
AIVQC Server
Publish approved versions only to selected stations.
- +Local deployment
- +Access control
- +Release history
03INSPECT
AIVQC Production
Make local OK/NOK decisions and pass them to the process.
- +Local image analysis
- +Recipes and rollback
- +Modbus TCP and history
HOW AN INSPECTION GOES LIVE
AI deployed in a controlled way.
A model is not the end of the project. Only an approved package, station qualification and unambiguous decision logic create a production solution.
- 01COLLECT
Collect images
Natural process variation and real OK and NOK examples.
- 02TRAIN
Prepare the model
Annotation, data versioning, training and result comparison.
- 03QUALIFY
Set criteria
Quality, errors and performance measured on an independent test.
- 04DEPLOY
Publish the package
The verified model goes only to selected stations.
- 05IMPROVE
Improve the process
History, difficult cases, a new version and safe rollback.
APPLICATIONS
Defects visible to a human.
Decisions made by the system.
A good image is essential for a good inspection. The best first application is a repeatable product, controlled position and a defect that can be shown clearly in an image.
PLASTIC HOUSING
Flash detected
NOKASSEMBLY
Component missing
NOKPRODUCT LABEL
Position correct
OKExample process scenarios — not reference validation results.
Missing component
Check assembly completeness and the presence of required components.
Incorrect assembly
Detect reversed, displaced or incorrectly seated parts.
Flash and deformation
Identify moulding, geometry and product-edge defects.
Cracks and damage
Visually detect scratches, cracks and surface damage.
Labels and print
Verify the presence, orientation and legibility of markings.
Contamination
Detect foreign objects, dirt and surface deviations.
PLASTICSAUTOMOTIVE TIER 2/3METAL COMPONENTSPACKAGING
DESIGNED FOR PRODUCTION
The line does not wait for the Internet.
AIVQC Production keeps the latest verified package locally. Inspection can continue during a server or network interruption, while activation and rollback remain controlled.
- 01 Local copy of verified packages
- 02 Permissions matched to the user's role
- 03 The system detects loss of inspection capability
- 04 Inspection history in a local database
PLANT NETWORK / ON-PREMISES
TRAINEREngineering workstationAPPROVED PACKAGE
SERVERLocal Docker / NASVERIFIED DELIVERY
PRODUCTIONEdge PC · Line 01 READY FOR OFFLINE OPERATION
EVIDENCE, NOT A PROMISE
Every inspection application receives its own validation result.
We do not transfer a demonstration result to your process. We measure the solution on representative data and under target imaging conditions.
QUALIFICATION REPORTGO / CONDITIONAL GO / NO-GO
- Product and conditions
- product type · camera · lighting
- Data
- image count · defect classes · test split
- Quality
- recall · precision · false accepts and rejects
- Performance
- decision time · throughput · target hardware
A reference validation result is not a guarantee for other products or process conditions.
AIVQC VS OTHER APPROACHES
A productised deployment process without being locked into one camera.
This comparison shows the typical scope of each approach. Final capabilities depend on the specific supplier and project.
| Capability | Manual inspection | Smart camera | Custom project | AIVQC |
|---|---|---|---|---|
| Automated OK/NOK decision | — | ✓ | ✓ | ✓ |
| AI-based defect detection | — | partly | ✓ | ✓ |
| Existing hardware qualification | — | — | depends | ✓ |
| Controlled model lifecycle | — | — | depends | ✓ |
| Local operation inside the plant | ✓ | ✓ | ✓ | ✓ |
| Pilot before deployment | — | depends | depends | ✓ |
| Scaling to additional lines | — | depends | depends | ✓ |
AIVQC specialises in quality inspection: from data and validation to local execution and version control across additional lines.
AIVQC PILOT
First, let's measure whether it works.
A paid feasibility pilot reduces risk before investment in a production station. You receive a data-based result — including when the honest recommendation is NO-GO.
PL
DEPLOYMENT AVAILABILITY · POLAND
Pilot and production deployments are currently available in Poland only. Future markets have not yet been selected.
STANDARD PRICE1,990 EURnet · one inspection application
Check whether we can detect your defect →- 01
Initial consultation
We define the product, defect, business value and imaging conditions.
- 02
Data
We prepare representative OK/NOK samples and an independent test set.
- 03
Model
We train up to three main iterations and select decision thresholds.
- 04
Test
We measure quality, false accepts, false rejects, performance and decision time.
- 05
Decision
You receive a report and a GO, CONDITIONAL GO or NO-GO recommendation.
FOUNDER-LEDINDUSTRIAL AI
MANUFACTURING EXPERIENCE + SOFTWARE
Built by someone who knows production and software.
AIVQC was created from practical experience in manufacturing, quality inspection, industrial maintenance and computer vision. The goal is not another AI demonstration, but a solution that can be qualified, maintained and improved under real plant conditions.
Dawid Oleśko · Founder, AIVQCFAQ
Frequently asked questions.
Every station and process is different. A pilot replaces assumptions with a real test.
Do my images have to go to the cloud?+
No. The AIVQC architecture is designed for local operation, and image analysis runs at the AIVQC Production station.
Can I use my existing camera or computer?+
Yes, if they pass qualification with the target camera, model, recipe, I/O and required cycle time. A processor or camera name alone does not guarantee the result.
Does the pilot guarantee a specific accuracy?+
We do not use one threshold for every project. Acceptance criteria are agreed before the pilot based on the risk and economics of the specific process.
What happens after a successful pilot?+
We prepare a separate production deployment offer covering the licence, hardware qualification, configuration and agreed integration scope.
SEE. DETECT. IMPROVE.
Let's talk about your process.
Describe the product, defect and current inspection method. We will make an initial assessment of whether the application is suitable for a pilot.
oleskodawid@aivqc.com- No-obligation response
- Data sent directly to AIVQC
- Initial inspection-application qualification