Computer Vision Anomaly Detection for Semiconductor Defect Classification

Period: Aug 2026 to May 2027
Type: NTU Final Year Project (SC4079 / SC4080)
Sponsor: Advanced Micro Devices (AMD)
Student: Ray Shao Yingzhan Selection: One of 4 students selected by AMD through an interview. Focus: building the anomaly detection model is now the main focus; the pipeline is secondary.

Project Scope & Objectives

Developing an end-to-end computer vision anomaly detection and defect classification pipeline targeting wafer testing and semiconductor quality assurance, building directly upon domain knowledge acquired during the AMD Datacenter GPU Product Development Internship.

Technical Focus Areas

• High-resolution microscopy wafer image processing and noise filtering.
• Unsupervised and semi-supervised anomaly detection architectures for rare defect identification.
• Automated defect clustering and yield impact telemetry integration.

Milestones