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.