A developer is building a smart kitchen shelf system using ESP32-S3 with an OV5640 fisheye camera to automate inventory tracking. The setup captures shelf images when the door opens, using edge AI (YOLOv8-lite) to recognize 50+ items with 92-98% accuracy. Data is sent via MQTT to a .NET Core backend for processing.

The backend, built with ASP.NET Core and ML.NET, uses a LangChain.NET agent to analyze inventory and generate meal suggestions based on family preferences. It can integrate with e-commerce APIs for auto-ordering. The project targets a single 80x40x60cm cabinet first, costing approximately 5044 RUB in hardware.

This prototype demonstrates a practical application of combining embedded AI with cloud services for home automation. The developer aims to expand to the full kitchen once the initial build is tested. Part 2 will cover real-world accuracy and code, moving from concept to a functional system.