AI on the Edge and Azure AI Platform - In person event

If disconnected and/or "real-time" processing is required, Edge AI is the only realistic solution.

However, Edge AI solutions can be hard to develop, test, install, update and support.

Key reasons for this include limited CPU power, hardware integration, unreliable connectivity, real-time processing requirements, hostile environments and physical tampering.

Another issue is "temporal coupling" and its impact on solution design.
There are often two parts to a successful solution e.g. processing of video with a lightweight model on the device (Edge AI) then uploading relevant images to Azure for post processing.

Edge AI often uses optimised models and specialised hardware like NVIDIA Jetsons with CUDA-TensorRT.

Azure Cloud AI is designed for speed of development, robustness, reliability, security, and scalability but also has some challenges like "bill shock".