Edge Artificial Intelligence (Edge AI) is a paradigm change in spatially distributing computational assets to process and infer with an applicable machine learning model locally on edge devices, including sensors, microcontrollers and embedded systems instead of using just cloud infrastructure. This has been resolved to provide an increasing demand of ultra-low latency, reduced bandwidth usage, increased energy efficiency, and a much security of data that cannot be ignored in application areas like autonomous vehicles, wearable healthcare devices, smart surveillance, and industrial automation. Since the complexity of inference tasks increases, it can be noted that the edge ceases to operate as a raw data path; instead, it actively contributes to intelligent decision-making. The Real-Time Operating System (RTOS) is in the centre of this change, where lightweight, but deterministic operating system layer controls the task execution, the use of memory, interactions with the available hardware within embedded systems. Unlike general-purpose operating systems, RTOS kernels are specifically designed to provide certain execution time and predictable execution time, which is the time embargoing necessity in AI workloads using neural network inference. Convolutional layers, activation functions and tensor operations thus should be closely coordinated under the constraints of real-time failure of mission-critical applications.