Accelerating Visual Test with Accessible AI for High-Accuracy Model Deployment

AI is transforming semiconductor test and inspection, yet widespread adoption is slowed by the complexity of integrating AI with existing workflows and legacy inspection tools. The Averroes.ai platform, built on an AI engine trained on millions of real-world defects, addresses this challenge by empowering test and process engineers to rapidly enhance and deploy visual inspection models without requiring a data science team.

Rather than replacing Automated Optical Inspection (AOI) systems, Averroes.ai enhances them by extending their capabilities with deep learning to detect subtle and novel defects, reduce false positives, and adapt to production variability.

This presentation will highlight how fabs and OSATs are using the platform to:

Train models with over 97% accuracy using existing production images

- Seamlessly integrate AI into legacy AOI and visual inspection systems

- Automatically label, fine-tune, and deploy models to test benches in under a week

- Leverage active learning and human-in-the-loop feedback to continuously improve performance

- Detect previously missed anomalies and reduce the manual classification burden

- Enable process and test engineers to iterate AI models directly on the production line without depending on data scientists

By building a robust, production-grade AI engine, the Averroes.ai platform transforms traditional inspection stations into intelligent, adaptive quality control systems. We will present case studies across frontend and backend processes, demonstrating measurable improvements in yield, responsiveness, and inspection accuracy.

This session is ideal for QA leads, test engineers, and yield teams looking to future-proof existing inspection equipment with AI. No costly infrastructure changes or steep learning curves are required.