Production ML for Mission-Critical Applications

Explore the challenges and best practices for deploying machine learning in mission-critical production environments. Delve into the rigorous approach required for maintaining and improving model performance over time, addressing issues unique to ML and data science. Learn about ML pipeline architectures, with a focus on Google's experience using TensorFlow Extended (TFX) for large-scale applications. Discover techniques for deep performance analysis, including edge and corner cases, as well as model sensitivity measurement. Gain insights into addressing software development methodologies, scalability, training/serving skew, and component modularity in ML applications. Understand the importance of comprehensive metrics beyond top-level performance, considering model fairness and predictive performance across user segments.