We show our work. Every number, every method, every time.
Insightek builds manufacturing AI with published test methods, auditable algorithms, and data that never leaves your factory. Two products — visual inspection and action compliance — one standard of proof.
Hero — dark factory line with overlaid detection bounding boxes and hand-tracking skeleton
Two products. One standard of proof.
Visual inspection and action compliance — both built on published test methods, auditable algorithms, and on-premise data sovereignty.
Visual Inspection AI Agent — product card image (PCB with semantic overlays)
Visual Inspection AI Agent
Identifies screws, capacitors, LEDs, labels — by understanding what they are.
Every metric published with its test method. Integration from AI to PLC in real time — the result hits your PLC before the operator lifts their hand.
See product →Action Compliance — product card image (hand-tracking skeleton over worker's hands)
Action Compliance & Bottleneck Analysis
An algorithm your process engineer can audit. Nine KPIs you can write into the contract.
Three weeks to live detection — including your engineer signing off every alert rule. Video never leaves your factory.
See product →Built for procurement, not just for demos.
Published methodology
Every number on our site has a test method and a comparison baseline. We disclose how we measured, not just what we measured.
Auditable algorithms
Our action compliance system uses a three-level scoring formula your process engineers can inspect, tune, and sign off on.
Data that stays inside your factory
Both products support full on-premise deployment. Images, video, and structured data never cross your network boundary.
Customers and ecosystem partners
- Customer 1
- Customer 2
- Customer 3
- Customer 4
- Customer 5
- Audi
- Apple
- Continental
- Bosch
- Siemens
Logos represent current customers and ecosystem collaborations. Specific engagement scope is available on request.
What changes after we deploy — and how we measured it.
Electronics assembly deployment
Same deployment, measured post-launch
Precision electronics assembly, action compliance rollout
| Metric | Value | Test method | Baseline |
|---|---|---|---|
| Engineering debug time | −95% | End-to-end time from blank line to first OK/NG for a new product | Prior traditional machine-vision setup at the same customer |
| Single-station throughput | +50% | Takt-time data, post-deployment vs. pre-deployment | Manual visual inspection baseline |
| Programming lines of code | 0 | Engineer timesheet during model registration | — |
Figures are from a named customer deployment under NDA. Specific numbers vary by line, product mix, and lighting. A full methodology report is available after NDA.
12 patents, honest status labels, and test methods you can verify.
We never use the word "certified" without the certificate. We never publish a number without the test method.
See technology & patents →Patent status sample
12 total- Multimodal feature fusion for industrial vision Granted
- Edge-cloud distributed inference scheduler Granted
- Semantic baseline from OK-sample learning Pending
- Three-level action localization Pending
Full patent list with authority, filing number, and status is available on the technology page.
See the test methods. Meet the engineering team.
Book a 30-minute demo. We'll walk through the methodology behind every claim on this page.