Edge deployment means running AI models close to where data is generated - on devices, local servers, or regional computing nodes - rather than sending everything to a centralised data centre. Use cases include autonomous vehicles processing sensor data in real time, factory floor systems inspecting products as they pass on a conveyor belt, retail cameras analysing foot traffic, and mobile apps running AI features without an internet connection. The benefits are compelling: lower latency (no round trip to the cloud), reduced bandwidth costs (process data locally rather than transmitting it), better privacy (data stays on the device), and resilience to connectivity issues. The challenges are equally real: edge devices have limited compute, memory, and power; updating models across thousands of distributed devices is operationally complex; and monitoring performance in production is harder when your infrastructure is scattered across many locations. Frameworks like TensorFlow Lite, ONNX Runtime, and Apple's Core ML help optimise and deploy models for edge environments. Model compression techniques - quantisation, pruning, knowledge distillation - are essential for fitting capable models into constrained devices. The trend toward more powerful edge hardware and more efficient models is steadily expanding what's possible at the edge, and many applications that required cloud processing a few years ago can now run locally.