How Amazon uses Amazon Nova models to automate operational readiness testing for new fulfillment centers

This article describes how Amazon uses Amazon Nova models to automate operational readiness testing (ORT) for new fulfillment centers, reducing manual verification time by 60%.

• ORT traditionally requires 2,000 manual hours per facility to verify 200,000+ components across 10,500 workstations
• Amazon Nova Pro selected for object detection with precise bounding box coordinates and high throughput
• Solution generates standardized UIN descriptions using Claude Sonnet to improve detection accuracy
• Serverless architecture uses Lambda, Bedrock, S3, and DynamoDB for cost-effective scalability
• Achieved 92% precision with 2-5 second latency per image in production testing
• System identifies missing components and improves ground truth data quality automatically
• Performance degrades for modules with 40+ components; hierarchical processing recommended

The IORA solution demonstrates how vision AI can automate visual inspection tasks across manufacturing, logistics, and quality assurance industries.