Objective 6.2

PMLE

Monitoring, testing, and troubleshooting AI solutions

Objective 6.2 sits in Monitoring AI solutions, which carries 13% of the ML Engineer exam. The questions below are original, written from the official objective title above, and each explanation cites the Google Cloud page it rests on.

Objective title verbatim from the official objectives. Google Cloud exam page

A worked example

Shown solved, with the whole explanation open: this is what every question here carries.

6-2Monitoring AI solutions

A team no longer holds the original training data and wants to catch changes in production inputs over time. Which detection applies?

Skew detection, using the training distributionThat comparison needs the training set you lack.
Attribution monitoring, using contribution scoresAttribution watches importance rather than distribution.
Drift detection, using past production dataCorrect · your answerCorrect. The reference is production history.
Evaluation, using a labelled holdout setLabels are not required for distribution monitoring.

Correct.

Concept

A distribution comparison needs two sides. Which comparison you are able to run is decided by which reference you still hold, so the absence of the training set picks the method for you.

Why C

Training-serving skew occurs when the feature data distribution in production deviates from the distribution used to train the model, and skew detection needs the original training data. Inference drift occurs when the production distribution changes significantly over time, and drift detection is used when the training data is unavailable.

Source

Training-serving skew occurs when the feature data distribution in production deviates from the feature data distribution used to train the model. If the original training data is available, you can enable skew detection to monitor your models for training-serving skew. Inference drift occurs when feature data distribution in production changes significantly over time. If the original training data isn't available, you can enable drift detection to monitor the input data for changes over time.

Introduction to Model Monitoring, Agent Platform documentation, checked August 2026
#gcp#model-monitoring#drift#skew

Now you: objective 6.2 questions

No account needed. The explanation opens when you answer.

Sample question 1 of 3

6-2Monitoring AI solutions

An engineer monitors one numerical feature and one categorical feature and asks which distance measure each uses. What is the pairing?

Sample question 2 of 3

6-2Monitoring AI solutions

A team changes nothing but expects the skew baseline to refresh as production data accumulates. When is that baseline recalculated?

Sample question 3 of 3

6-2Monitoring AI solutions

A team enables monitoring on a model and names neither features nor thresholds. What does the default configuration cover?

Full ML Engineer question bank coming

We’re writing the complete bank from the official objectives right now. Leave your email and we’ll tell you when it ships, nothing else, ever.

Read the sources

These are the official pages the questions above cite. Reading them is studying the objective from the primary source, which is what the explanations point you toward anyway.

More objectives in Monitoring AI solutions