← All case studies

AI/ML · Broadcasting · Automation

AI Sports Live-Streaming Platform

A camera-system integration platform combining AI-driven video analysis and event tracking for automated sports broadcasting.

RoleTest Automation Engineer
TimeframeMar 2024 - Jul 2024
TeamIsraeli Company

My role

  • Owned the mobile-automation direction within a 5–6 person QA team; I was the only engineer working on this front.
  • Extended an existing Python, Robot Framework and Appium suite for a mobile app connected to real camera hardware over cable and Wi-Fi.
  • Also contributed BrowserLibrary automation for another part of the platform, alongside test execution, debugging and defect reporting.

Technical challenges

  • The mobile app and physical camera could disagree about their connection or readiness state, so UI state alone was not sufficient evidence that a workflow could proceed.
  • Some mobile interactions required screen-coordinate navigation rather than ordinary Appium element selection.
  • The tests covered workflows spanning device setup, camera communication, recording or broadcast state, and video availability.

My approach

I extended the existing Robot Framework and Appium suite with mobile workflows for connecting or pairing a camera, starting and stopping recordings or broadcasts, selecting a sport or event, changing camera or stream settings, viewing live or recorded video, and managing accounts. I also added supporting code around device setup and randomized input.

I added roughly 30–50 automated mobile tests against real cameras and integrated them into the project's larger CI regression suite. The coverage exposed state and media mismatches, including cases where the camera showed one image while the app displayed another, as well as controls enabled before the camera was ready.

Results

30–50
Appium tests added
~18 hrs
CI regression runtime
~2 weeks
Manual regression replaced
  • The 30–50 Appium tests added repeatable mobile-to-camera coverage to CI for workflows that previously depended heavily on manual hardware testing.
  • The full regression suite ran in sections over roughly 18 hours, replacing approximately two weeks of manual regression across the wider product.
  • Camera/app state and media mismatches were fixed and successfully retested. The client accepted the automation work for the already-released product's ongoing development.

Stack

PythonRobot FrameworkAppiumBrowserLibrary

More case studies