AI cloud data platform
Inspection data gathered at the edge is analysed together — defect trends by line, product and period on one screen. Cloud deployment is the customer’s choice.
Inspection does not end at a pass/fail verdict. As images and judgements accumulate from the line, the relationship between process conditions and root causes becomes visible. DSTEK supplies the data layer from edge to cloud, closing the loop from defect detection through root-cause analysis to process optimisation. The default deployment is on-premise; cloud connectivity is configured only when the customer asks for it.
- Database scale
- Petabyte-class image DBMS
- Architecture
- Integrated edge and cloud
- Integration
- Real-time MES and PLC
- Model
- Industrial vision foundation model
Data Intelligence
From detection to autonomous decisions
Four stages forming a single closed loop.
- 01 Detection Real-time judgement on the line
- 02 Accumulation PB-scale image and verdict history
- 03 Root cause Correlating process conditions with defects
- 04 Optimisation Condition adjustment and retraining from feedback
Data Intelligence
Components
AI data traceability
A petabyte-class image database with full history, so the images and reasoning behind a given product’s verdict can be retrieved after the fact.
Industrial vision foundation model
A base model trained on accumulated industrial imagery, shortening initial training for new products and serving as the starting point for transfer learning.
Real-time monitoring
Model performance and inspection quality are watched while the line runs, flagging when drift means retraining is due.
Data Intelligence
Why it compounds
The reason the system is worth more the longer it runs.
Field data
2B+/day
Images analysed daily across the DSTEK group.
Traceability
PB+
Inspection images and reasoning retrievable per product.
Retraining
Drift detection
Performance decay caught early, so retraining happens when it should.
Data security
Inspection data stays inside your facility
The default deployment is on-premise: training, inference and storage all run on customer infrastructure, and cloud connectivity is configured only on request.