AI Security System with Mobile Notifications -- 2

I’m building an AI-driven security solution that watches live video feeds, decides in real time whether something needs my attention, and instantly pushes a mobile notification when it does. The heart of the job is the end-to-end pipeline: ingest the camera stream, run fast and reliable AI inference, and fan out alerts to my iOS and Android app with minimal delay. Every alert should arrive with a snapshot, timestamp, and location tag so I can act immediately from my phone.

Low latency, high accuracy, and rock-solid reliability matter more to me than bleeding-edge experimentation. I’m happy for you to choose the framework—TensorFlow, PyTorch, OpenCV, or another stack—as long as the result is well-documented, container-friendly (Docker/Kubernetes welcome), and easy to scale from one site to many.

Deliverables
• A working prototype that detects events and sends push notifications in under five seconds.
• Full source code plus clear setup instructions.
• A brief test report covering detection accuracy and notification speed.

Acceptance criteria: during controlled testing, the system must deliver a mobile push alert within five seconds of a relevant event and average no more than one false alert per hour.