Inspection keeps getting more sensitive, and every new node or bonding step brings defects nobody has classified before. When the evidence behind a call is thin, operators play it safe, and good wafers end-up scrapped.
Robovision turns inspection signals into ranked, auditable evidence, in the fab's review flow or inside the process tool. You take the decisions, we supply the evidence behind them.
Indurion classifies and measures each defect on every wafer, so lots no longer get judged from a small sample. The same defect gets the same verdict, whatever the shift or reviewer.
As defects change, drift monitoring flags data the models haven't seen before, and the models get retrained. The same decision logic carries across your tools and sites.
Local inference is offered by others. But I couldn’t find a single other industrial AI vision intelligence platform that let us do the labeling, the training and everything else locally on the machine, fully air-gapped.
Higher yield: keep the good wafers you scrap today
In a representative high-volume fab, sample-based review costs an indicative ±€3.8M per year in avoidable scrap and review effort. On a hybrid-bonding line, false scrap and bond failures add up to around ±€3M per year.
For tool builders, that evidence becomes part of the machine, under their own brand: a tool that contributes to customer yield, with a recurring software margin on top of the hardware.
Fits the tools already in place
Indurion is hardware-agnostic and runs at the edge. It reads standard inspection output such as KLARF files and wafer maps, so inspection hardware and machine controls stay as they are.
New defect types, same application
When a new node or bonding step brings unfamiliar defects, engineers add and validate the new defect classes in the running application instead of starting a new project.
Every change remains traceable
Each model version and change is logged. If a new version underperforms, the line rolls back to the last validated one in under 30 minutes.
Your wafer data stays yours
Production data trains your own models only. It is not pooled with data from other companies. Models and defect libraries stay under your control as you scale.
Measured against your payback case
With Value Assurance, Robovision's Services measure a baseline such as scrap rate or review effort, then track yield recovery against the payback scenario. If a target is missed, next steps are defined.
Robovision's services stay with your team from deployment through daily production. Support levels are agreed around the tools and process steps that matter most to your yield.
Value Assurance tracks recovery against your payback case. For OEMs, SLA-defined support and predictive maintenance add recurring service revenue across the installed base.
Inspection now flags more anomalies than reviewers can classify with confidence. See why uncertain calls on a few sampled wafers drive lot-level scrap, and how evidence-backed review can recover millions of euros per fab, per year.
See what inconsistent review and classification uncertainty could be costing your fab. Use your own workflow assumptions to model recoverable wafers and the annual financial upside of full-coverage, decision-backed review.
Discover readiness across imaging, compute resources, control-loop integration, and customer context. In minutes, see how close your platform is to turning inspection data into governed, real-time decisions at the point of action.
Find answers to common questions about deploying and governing vision AI in semiconductor production. Still need help? Please get in touch with our team.
Yes. Indurion is hardware-agnostic and runs on x86/64 architecture. It can integrate with existing imaging and compute infrastructure, provided the installation and performance requirements are met.
You do. The models and application IP you develop remain yours, giving you control over the AI capabilities you build.
Indurion gives teams the infrastructure to deploy, manage, scale, and govern vision AI applications in production. It reduces the manual work involved in maintaining applications over time, while your teams retain control over how they are configured and deployed.
Yes. Teams can update and validate models as requirements change without rebuilding the application from scratch, reducing the specialist intervention required to keep vision AI current.
Indurion Agent, the component that runs vision AI applications locally at the edge, can continue inference without a continuous network connection.
Yes, deployed versions and changes are tracked throughout the application lifecycle. Teams have the option to roll back to a previous validated version when needed.
Robovision provides services to support your team from implementation through production. We offer additional expertise or capacity where needed, as well as value assurance to measure performance against the expected business case over time. For more info check the Services section
We’ll help you identify where vision AI can create the most value.
