No Cloud, No Latency, Global Reach, Local Processing

Testing streaming applications at scale means juggling regional differences, device variety, network conditions, cloud costs, and strict privacy requirements, often with data that simply can’t leave its location.

This session introduces an edge-first approach to automated QA: a computer vision pipeline running entirely on a compact edge device with a dedicated neural inference accelerator. Deploy it anywhere in the world, manage it remotely, and run visual tests without sending screenshots, recordings, or UI data outside the target environment.

The architecture features a multi-model routing mechanism that detects the active streaming application and dynamically loads the right YOLO model for UI component recognition — all processed locally, with full infrastructure control retained centrally.
We’ll also cover the real work behind efficient edge AI: dataset preparation, model training, quantisation, deployment on neural accelerators, and the trade-offs between accuracy, latency, and resource use.

The result: a distributed QA platform that is fast, secure, and fully cloud-independent with global reach and local processing.