Model monitoring

Tracking a deployed model's inputs, outputs, and performance over time in production, distinct from infrastructure monitoring like uptime or CPU usage. A model can stay perfectly available while its predictions quietly get worse, which is the failure monitoring exists to catch.

Why exams ask this

Tested as a "which metric should trigger the alert" scenario. The distractor picks an infrastructure signal such as uptime or request latency, when the correct trigger is a shift in the prediction or output distribution: infrastructure health and prediction quality are measured by entirely different signals.

Relevant to

Related concepts

Resources

No resources linked to this concept yet.