As artificial intelligence (AI) transforms industries, state-of-the-art models have exploded in size and capability. However, deploying these models on resource-constrained edge devices remains a significant challenge. Smartphones, wearables, and IoT sensors face stringent limitations on compute, memory, power, and communication, creating a gap between demanding AI models and edge hardware capabilities that hinders the deployment of intelligence. In this talk, we will re-examine techniques to bridge this gap and embed big AI on small devices. We will begin by discussing how the properties of various hardware platforms impact the design strategies of efficient deep neural network (DNN) models, such as quantization and pruning. Next, we will discuss techniques aimed at reducing the inference and training costs of distributed collaborative edge AI systems. Finally, we will delve into the underlying design philosophies and their evolution toward efficient, scalable, robust, and secure edge computing systems.