DevFest Seattle 2022: Common questions on ML Observability: why, when, how

One of the key parts of the MLOps discipline is observability - practice of instrumenting ML systems to gather actionable data for monitoring, troubleshooting and creating feedback loops. We will discuss the core components of an observability solution designed specifically for ML pipelines. The talk will dive into different approaches of instrumenting models and pipelines, collecting actionable telemetry and catching model failures that are caused by data/concept drifts and/or data quality issues.

Speaker: Alessya Visnjic, WhyLabs