Why Model Observability Is the Next Data Engineering Frontier

What this episode covers

Lucas and Luna dive into the growing field of model observability—how companies monitor machine learning models in production beyond just accuracy metrics. They discuss the 2025 Aporia/WhyLabs survey showing 72% of enterprises have suffered a model-degradation incident costing over $200,000, and why traditional data observability tools miss ML-specific issues like data drift, concept drift, and feature skew. The episode centers on a case study: how a mid-size e-commerce company caught a 15% revenue drop from a model that silently retrained on corrupted data, saved by real-time drift detection. They explore the emerging stack: WhyLabs, Arize AI, Evidently AI, and the shift from batch monitoring to streaming observability. Lucas argues that as ML models become more embedded in core business logic, observability is shifting from a data-engineering concern to a boardroom priority. Luna questions whether the tooling is mature enough for non-tech enterprises. The episode closes with a reflection on the cost of not knowing what your model is doing. #ModelObservability #MLMonitoring #DataDrift #ConceptDrift #MachineLearning #DataEngineering #ArizeAI #WhyLabs #EvidentlyAI #Aporia #FeatureSkew #ProductionML #DataQuality #BusinessAndTechnology #FexingoBusiness #BusinessPodcast #TheDataBusinessPodcast #DataInfrastructure Keep every episode free: buymeacoffee.com/fexingo

PodParley-generated summary based on available episode metadata and transcript content.