AI System Validation - Meta Perspective: The Challenges of Debugging at Scale

Learn about the challenges and approaches to debugging at scale in this 17-minute talk by Carlos Fernandez, an AI System Validation Engineer at Meta. Explore the complexities of debugging highly interconnected systems, managing the risks of constant updates in a fast-paced development environment, and addressing diverse use cases across AI, Storage, and Compute. Discover Meta's effective strategies including real-time fleet monitoring with alert-triggered repairs, automated testing frameworks like formal Firmware Qualification to catch issues early, and cross-functional collaborative debugging that leverages diverse expertise to solve complex problems.