Edge computing refers to processing data at or near the source of data generation, rather than sending it to centralized servers or cloud infrastructure. The goal is to reduce latency, save bandwidth, improve responsiveness, and enable real-time decision-making—especially in scenarios where immediate feedback is critical, like autonomous vehicles, industrial automation, and IoT systems.

By performing computation locally on devices such as sensors, cameras, or embedded hardware (e.g., NVIDIA Jetson, Raspberry Pi), edge computing reduces dependency on stable internet connections and cloud infrastructure. This is useful for privacy-sensitive applications, bandwidth-constrained environments, or systems that require high availability.

For machine learning, this means running models directly on edge devices. Techniques like model quantization, pruning, and knowledge distillation are used to make models lightweight enough for edge deployment.

Edge computing is increasingly paired with cloud systems in hybrid architectures, where the edge handles fast, local tasks and the cloud handles heavier processing or long-term storage.