Lightning Talk: Machine Learning Observability with Prometheus
summary
Shows how to use Prometheus + Grafana to monitor ML models in production by scraping model endpoints and short‑lived jobs, storing samples, running rules to aggregate metrics and fire alerts. Tracks concrete signals: latency and resource metrics (CPU, memory, disk), plus prediction distribution details like min/max, standard deviation and drift over time. Includes a dashboard example for average house price predictions and explains using Prometheus rules to generate alerts on metric regressions.
AI-written summary · about
chapters (4)
related talks
24mObservability in the MLOps Lifecycle with Prometheus
37mWhat I learned from monitoring more than 30 Machine Learning Use Cases
01h 12mDetecting Workload Anomalies with Prometheus and Machine Learning
37mLowering the barrier for ML monitoring
25mSymptom-based Alerting for Machine Learning - What I Learned...
50mMetrics and alerting with Prometheus and Grafana
09mGraduated Project Lightning Talk: Using Trace Data for Monitoring & Alerting of Appl...
24mFrom Monitoring to Observability: 10 Things to Do with Prometheus in 2020