Revolutionizing Testing at Uber

Overview

In this episode of the Mobile Reliability Podcast, Anam Hira talks with Uber engineer Juan Marcano about how DragonCrawl, an AI-powered testing system, transformed mobile testing at Uber. Juan shares how he stumbled into software engineering, then into machine learning through projects and open source, and why traditional manual and scripted testing couldn’t keep up with a global, fast-moving app.

They dig into how DragonCrawl uses LLMs to explore the app, run flows, and keep tests stable, cutting maintenance costs, streamlining painful localization testing, and helping quantify revenue leakage in the millions from bugs. The result: a testing system that pushed critical test pass rates to 99.3% and became a major productivity boost for Uber’s developers.

Key Topics

• Juan’s journey into software engineering and machine learning
• Why traditional testing couldn’t scale at Uber
• How DragonCrawl uses LLMs to explore and test the app
• Reducing maintenance costs and stabilizing tests
• Localization testing challenges and solutions
• Quantifying revenue leakage from bugs
• Achieving 99.3% critical test pass rates