Symptom-based Alerting for Machine Learning - Lessons from 30+ Use Cases

Explore symptom-based alerting for machine learning in this 25-minute conference talk from SREcon23 Europe/Middle East/Africa. Discover practical insights on implementing effective monitoring for ML stacks, going beyond traditional software monitoring practices. Learn which metrics to prioritize, how to detect issues in real-time, and whether existing tools suffice or if an MLOps platform is necessary. Gain valuable knowledge from the speaker's experience monitoring over 30 machine learning use cases, equipping you with strategies to enhance your ML monitoring capabilities.