Robovision

The tool flags the anomaly. How is it classified as killer, nuisance, borderline, or acceptable? The classification may come from a static rule, a local script, a closed third-party system, or a person reviewing a sample. Each route can introduce a different control gap, while the production decision may still sit with the customer’s systems or people. Robovision’s Indurion Suite embeds vision intelligence into inspection and process tools. It turns detections into consistent classifications and structured decision evidence, delivered as part of the OEM’s machine. Follow one anomaly from detection to decision evidence.

Robovision’s Indurion Suite embeds vision AI into inspection and process tools. It enables detections to be interpreted within the machine and used to support the next decision in the workflow. The capability becomes a governed, repeatable part of the OEM’s product, deployable across machines and customer sites as requirements change. See where it sits between detection and the yield decision.

How much of the yield number measures quality, and how much measures caution? A reviewer sees only a fraction of what gets flagged. The resulting decision can still affect the whole lot, including wafers no one reviewed. Under incomplete evidence, the resulting disposition may be more conservative. Robovision enables consistent classification beyond the limits of manual review, giving fabs structured evidence to support lot-level decisions. The manufacturing execution system retains final control over disposition. Follow one flagged defect from inspection to the annual yield number.

Robovision makes vision AI software that classifies what semiconductor inspection tools find. Each flagged anomaly is interpreted consistently across every tool and lot, while the manufacturing execution system keeps the final say on disposition. See why consistent classification matters to yield.

Your tool flags an anomaly. The customer still needs a yield decision. The hardware has done its job. The challenge comes when that signal moves into disconnected rules, local scripts, and review queues. Each new site adds another place for the decision to vary, while the engineering load grows with the installed base. Robovision’s Indurion Suite embeds vision intelligence into inspection and process tools. It enables detections to be interpreted consistently within the workflow, as part of the OEM’s machine. See what happens between anomaly detection and the yield decision.

What happens when a lot-level decision rests on a fraction of the evidence? Human review can cover only a fraction of flagged anomalies. But the resulting evidence still informs a decision on the whole lot. Under that uncertainty, scrap is the safe call. In a modelled fab scenario, reducing that uncertainty represents around €3.8M in annual recoverable value. The loss comes from the decision after detection. Robovision enables consistent classification beyond the limits of manual review, giving fabs structured evidence for lot-level decisions. See how review decisions affect yield.