Skip to main content

Model Monitoring Drift Detection

Set up continuous monitoring for deployed models including data drift detection, concept drift detection, performance degradation alerts, and automated retraining triggers.

Complexity: Complex | Duration: 30m+ | Category: Devops

Tags: ml monitoring drift-detection mlops

Workflow Diagram​

Steps​

Step 1: Baseline capture​

Agent: ml

engineer - Capture training data statistics (mean, std, distributions)

Step 2: Data drift detection​

Agent: ml

engineer - PSI, KL divergence, Kolmogorov-Smirnov tests

Step 3: Concept drift detection​

Agent: ml

engineer - Monitor prediction distribution shifts

Step 4: Performance monitoring​

Agent: ml

engineer - Track accuracy, latency, throughput metrics

Step 5: Alert configuration​

Agent: devops

engineer - Set thresholds for drift and performance degradation

Step 6: Visualization dashboard​

Agent: devops

engineer - Grafana/Tableau dashboards for metrics

Step 8: Incident response​

Agent: devops

engineer - Document escalation and remediation procedures

Usage​

To execute this workflow:

/workflow devops/model-monitoring-drift-detection.workflow

See other workflows in this category for related automation patterns.