Shelf monitoring system with three-tier architecture separating presentation, application, and data layers across public and private subnets. User devices connect to a frontend server in the public subnet, which routes requests to backend and AI servers in private subnets for processing and inference. Cameras feed real-time shelf data to the AI server for inventory detection, with results persisted in a database server isolated in the data tier. This architecture enforces network segmentation and least-privilege access, critical for retail environments handling sensitive inventory and customer data. Fork and customize this diagram on Diagrams.so to adapt subnet configurations, add load balancers, or integrate additional camera feeds.