Yash Varyani

Locator-based testing frameworks like Espresso and Appium have long been staples in the Android developer’s toolkit—but they’re increasingly showing their age. As developers build more dynamic, visually complex, and AI-driven apps, these tools struggle to keep up. Tests break frequently, maintenance becomes a chore, and certain UI environments—like games, maps, or multimedia-rich interfaces—remain virtually untestable. In this talk, we’ll explore a new frontier in mobile testing: one powered by computer vision and large language models (LLMs). Instead of relying on brittle resource IDs or view hierarchies, we’ll look at how AI agents can test apps the way a human would—by seeing and interacting with pixels on screen. You’ll learn how visual object detection, LLM-based command resolution, and multimodal agent architectures are reshaping what’s possible in UI automation. Whether you’re tired of maintaining flaky tests or looking to future-proof your testing strategy, this session will give you a glimpse into the next generation of resilient, intelligent Android UI testing.