Google Professional Machine Learning Engineer Practice Test 7
Question 60
During a postdeployment AI safety and reliability review, the following condition occurs: A company assistant follows instructions embedded in an untrusted retrieved webpage, overriding the system-approved task and attempting to reveal internal data. The model owner must provide repeatable evidence rather than a cosmetic dashboard change. Which diagnostic check should be performed FIRST to isolate the described cause?
- Inspect attachment ingestion, trust boundaries, malicious prompt examples and deployed Model Armor filter results.
- Inspect retrieved document boundaries, instruction hierarchy, malicious test prompts and the application's output actions.
- Review prompt token counts, output length, cache behavior, p95 model latency and cost per successful task.
- Review the business safety taxonomy, representative adversarial evaluation prompts, safety-filter configuration and escalation design.
Explanation
Correct Answer
B:Inspect retrieved document boundaries, instruction hierarchy, malicious test prompts and the application's output actions. This inspects evidence relevant to prompt injection through retrieved documents. External retrieval content can carry attacker instructions; data should not receive the authority of trusted application instructions. After identifying the cause, the matching correction is: Treat retrieved text as untrusted data, isolate instructions, restrict tool actions and apply appropriate Model Armor screening.
Incorrect Answers
A:Inspect attachment ingestion, trust boundaries, malicious prompt examples and deployed Model Armor filter results. This examines model armor untrusted content screening rather than the specific mechanism reported here. It is not the decisive first check for prompt injection through retrieved documents. The relevant evidence is: Inspect retrieved document boundaries, instruction hierarchy, malicious test prompts and the application's output actions.
C:Review prompt token counts, output length, cache behavior, p95 model latency and cost per successful task. This examines llm latency and cost token regression rather than the specific mechanism reported here. It is not the decisive first check for prompt injection through retrieved documents. The relevant evidence is: Inspect retrieved document boundaries, instruction hierarchy, malicious test prompts and the application's output actions.
D:Review the business safety taxonomy, representative adversarial evaluation prompts, safety-filter configuration and escalation design. This examines gen ai output safety policy enforcement rather than the specific mechanism reported here. It is not the decisive first check for prompt injection through retrieved documents. The relevant evidence is: Inspect retrieved document boundaries, instruction hierarchy, malicious test prompts and the application's output actions.
Question 61
During a staged restoration test at a secondary serving deployment, the following condition occurs: A customer-facing inference API has high end-to-end latency in Asia while model execution is quick and the endpoint runs on another continent. The team must prove an actual user-visible improvement without weakening access controls. Select TWO: the decisive diagnostic step and the targeted repair for this specific issue.
- Inspect feature-view sync status, latest source timestamps, entity ID and online lookup freshness.
- Pin releases to reviewed immutable image digests and record image plus model artifact versions.
- Decompose client network, TLS, request transport, model compute and response times by user geography.
- Enable compatible dynamic batching with a bounded wait time and benchmark throughput against p95 latency.
- Select supported regional placement or routing close to users while respecting residency and model availability.
Explanation
Correct Answers
C:Decompose client network, TLS, request transport, model compute and response times by user geography. This inspects evidence relevant to serving region proximity versus model compute. Network distance can dominate inference response times when the actual model execution is already short. After identifying the cause, the matching correction is: Select supported regional placement or routing close to users while respecting residency and model availability.
E:Select supported regional placement or routing close to users while respecting residency and model availability. This is the targeted corrective response for serving region proximity versus model compute. Network distance can dominate inference response times when the actual model execution is already short. Verify the diagnosis using: Decompose client network, TLS, request transport, model compute and response times by user geography. Confirm success using: Regional clients experience lower p95 end-to-end latency without changes to model compute quality.
Incorrect Answers
A:Inspect feature-view sync status, latest source timestamps, entity ID and online lookup freshness. This is an action or diagnostic relevant to feature store freshness and online sync, not one of the required two steps. In this scenario verify: Decompose client network, TLS, request transport, model compute and response times by user geography. Then correct: Select supported regional placement or routing close to users while respecting residency and model availability.
B:Pin releases to reviewed immutable image digests and record image plus model artifact versions. This is an action or diagnostic relevant to model artifact image digest reproducibility, not one of the required two steps. In this scenario verify: Decompose client network, TLS, request transport, model compute and response times by user geography. Then correct: Select supported regional placement or routing close to users while respecting residency and model availability.
D:Enable compatible dynamic batching with a bounded wait time and benchmark throughput against p95 latency. This is an action or diagnostic relevant to dynamic batch serving for gpu saturation, not one of the required two steps. In this scenario verify: Decompose client network, TLS, request transport, model compute and response times by user geography. Then correct: Select supported regional placement or routing close to users while respecting residency and model availability.
Question 62
While preparing a production training workflow on Gemini Enterprise Agent Platform, the following condition occurs: A hospital alert model achieves 99% accuracy by predicting every patient is negative when the positive condition occurs in 1% of cases. Engineers need an auditable, resource-aware correction to the issue. Which evidence most directly establishes what is wrong with this AI or ML workflow?
- Examine model governance constraints, feature sensitivity, explanation fidelity and performance of simpler interpretable baselines.
- Inspect feature computation cutoffs, window frame bounds, label timestamps and time-aware evaluation splits.
- Inspect output schema, decoding configuration, structured-output support and malformed samples under representative prompts.
- Inspect prevalence, confusion matrix, false-negative harm, calibration and precision-recall curves at candidate thresholds.
Explanation
Correct Answer
D:Inspect prevalence, confusion matrix, false-negative harm, calibration and precision-recall curves at candidate thresholds. This inspects evidence relevant to cost-sensitive threshold for rare positives. Accuracy on highly imbalanced labels can hide severe minority-class recall failures. After identifying the cause, the matching correction is: Choose a decision threshold using acceptable recall and alert burden with a validated independent cohort.
Incorrect Answers
A:Examine model governance constraints, feature sensitivity, explanation fidelity and performance of simpler interpretable baselines. This examines interpretability-constrained model baseline rather than the specific mechanism reported here. It is not the decisive first check for cost-sensitive threshold for rare positives. The relevant evidence is: Inspect prevalence, confusion matrix, false-negative harm, calibration and precision-recall curves at candidate thresholds.
B:Inspect feature computation cutoffs, window frame bounds, label timestamps and time-aware evaluation splits. This examines temporal training label leakage in rolling windows rather than the specific mechanism reported here. It is not the decisive first check for cost-sensitive threshold for rare positives. The relevant evidence is: Inspect prevalence, confusion matrix, false-negative harm, calibration and precision-recall curves at candidate thresholds.
C:Inspect output schema, decoding configuration, structured-output support and malformed samples under representative prompts. This examines constrained generative json outputs rather than the specific mechanism reported here. It is not the decisive first check for cost-sensitive threshold for rare positives. The relevant evidence is: Inspect prevalence, confusion matrix, false-negative harm, calibration and precision-recall curves at candidate thresholds.
Question 63
During a regulated MLOps pipeline deployment, the following condition occurs: A scheduled training pipeline fails during feature validation but the team discovers the failure only when customers complain about stale predictions. The workflow must be repeatable and protect the last approved production model. Which TWO complementary choices jointly establish the evidence and appropriate correction? Select TWO.
- Train only on adequately matured labeled cohorts and monitor recent feature drift separately until outcomes arrive.
- Inspect task execution states, failed-job logs, alert routing, pipeline-age metrics and owner escalation settings.
- Inspect request schema, container route behavior, test fixtures and CI stage dependency graph.
- Publish one versioned pipeline package and promote it with reviewed environment-specific parameters.
- Configure failure and stale-model alerts tied to actionable run-status and freshness service objectives.
Explanation
Correct Answers
B:Inspect task execution states, failed-job logs, alert routing, pipeline-age metrics and owner escalation settings. This inspects evidence relevant to pipeline notification alert on silent failure. Orchestration without failure notification can leave models stale even when the last successful deployment was correct. After identifying the cause, the matching correction is: Configure failure and stale-model alerts tied to actionable run-status and freshness service objectives.
E:Configure failure and stale-model alerts tied to actionable run-status and freshness service objectives. This is the targeted corrective response for pipeline notification alert on silent failure. Orchestration without failure notification can leave models stale even when the last successful deployment was correct. Verify the diagnosis using: Inspect task execution states, failed-job logs, alert routing, pipeline-age metrics and owner escalation settings. Confirm success using: An injected validation failure alerts the responsible owner and the last healthy model continues serving.
Incorrect Answers
A:Train only on adequately matured labeled cohorts and monitor recent feature drift separately until outcomes arrive. This is an action or diagnostic relevant to scheduled retraining on delayed ground truth, not one of the required two steps. In this scenario verify: Inspect task execution states, failed-job logs, alert routing, pipeline-age metrics and owner escalation settings. Then correct: Configure failure and stale-model alerts tied to actionable run-status and freshness service objectives.
C:Inspect request schema, container route behavior, test fixtures and CI stage dependency graph. This is an action or diagnostic relevant to continuous delivery integration contract tests, not one of the required two steps. In this scenario verify: Inspect task execution states, failed-job logs, alert routing, pipeline-age metrics and owner escalation settings. Then correct: Configure failure and stale-model alerts tied to actionable run-status and freshness service objectives.
D:Publish one versioned pipeline package and promote it with reviewed environment-specific parameters. This is an action or diagnostic relevant to pipeline environment parameter promotion, not one of the required two steps. In this scenario verify: Inspect task execution states, failed-job logs, alert routing, pipeline-age metrics and owner escalation settings. Then correct: Configure failure and stale-model alerts tied to actionable run-status and freshness service objectives.
Question 64
At a Google Cloud analytics team preparing a governed pilot, the following condition occurs: A reporting team exports millions of BigQuery rows to a laptop for daily scoring although a trained BigQuery ML model already exists in the dataset. The team must satisfy a reproducible low-code AI assessment. Which statement identifies the primary reason for this failure?
- In-database prediction avoids unnecessary bulk data transfer when model and features are already in BigQuery.
- Clustering segments observations without learning their known outcomes, while supervised classification predicts those outcomes.
- Forecasting requires time-ordered histories and explicitly defined future steps, not shuffled categorical labels.
- Frequently changing external knowledge is better supplied through current context than repeatedly embedded into model weights.
Explanation
Correct Answer
A:In-database prediction avoids unnecessary bulk data transfer when model and features are already in BigQuery. This explains the mechanism behind bigquery ml sql batch prediction integration and why the described failure is possible. A discriminating investigation is: Review row volume, model location, permissions and the current SQL report dependency on the local scoring script. The corresponding repair is: Use ML.PREDICT on the appropriate BigQuery ML model within a controlled SQL batch job, writing results to governed tables.
Incorrect Answers
B:Clustering segments observations without learning their known outcomes, while supervised classification predicts those outcomes. This principle describes bigquery ml logistic model for conversion probability, which is not the cause of the observed bigquery ml sql batch prediction integration failure. The controlling explanation is: In-database prediction avoids unnecessary bulk data transfer when model and features are already in BigQuery.
C:Forecasting requires time-ordered histories and explicitly defined future steps, not shuffled categorical labels. This principle describes bigquery arima_plus forecast horizon, which is not the cause of the observed bigquery ml sql batch prediction integration failure. The controlling explanation is: In-database prediction avoids unnecessary bulk data transfer when model and features are already in BigQuery.
D:Frequently changing external knowledge is better supplied through current context than repeatedly embedded into model weights. This principle describes foundation model tuning versus retrieval update, which is not the cause of the observed bigquery ml sql batch prediction integration failure. The controlling explanation is: In-database prediction avoids unnecessary bulk data transfer when model and features are already in BigQuery.
Question 65
At a Google Cloud analytics team preparing a governed pilot, the following condition occurs: A support assistant includes a full 200-page manual in every prompt despite users asking brief questions, producing excessive latency and cost. The team must satisfy a reproducible low-code AI assessment. What acceptance result verifies the intended behavior after remediation?
- Held-out metrics reflect only features available before the prediction timestamp and remain credible on later periods.
- On held-out questions, answers remain grounded while input tokens and p95 response latency decline.
- Representative photographs return the needed detections at acceptable precision, latency and per-image cost.
- Classification F1 remains above the acceptance threshold while end-to-end latency and cost per thousand requests fall.
Explanation
Correct Answer
B:On held-out questions, answers remain grounded while input tokens and p95 response latency decline. This is specific acceptance evidence for gemini prompt context token budget. Unfocused context consumes tokens and can distract generation; targeted retrieval can reduce cost without discarding needed facts. It confirms the controlled correction: Retrieve only relevant passages with provenance and use a concise structured prompt with measured token limits.
Incorrect Answers
A:Held-out metrics reflect only features available before the prediction timestamp and remain credible on later periods. This result could demonstrate success for bigquery ml ml.evaluate leakage detection but not for the stated gemini prompt context token budget issue. The direct test is: On held-out questions, answers remain grounded while input tokens and p95 response latency decline.
C:Representative photographs return the needed detections at acceptable precision, latency and per-image cost. This result could demonstrate success for vision api pretrained image labels versus custom detection but not for the stated gemini prompt context token budget issue. The direct test is: On held-out questions, answers remain grounded while input tokens and p95 response latency decline.
D:Classification F1 remains above the acceptance threshold while end-to-end latency and cost per thousand requests fall. This result could demonstrate success for gemini model choice for latency versus complexity but not for the stated gemini prompt context token budget issue. The direct test is: On held-out questions, answers remain grounded while input tokens and p95 response latency decline.
Question 66
During a production ML inference release on Google Cloud, the following condition occurs: A document classification service sends huge media files through synchronous prediction calls whose processing exceeds clients' timeout limits. The service must meet its agreed quality and latency requirements. Which statement identifies the primary reason for this failure?
- Network distance can dominate inference response times when the actual model execution is already short.
- Mutable nicknames do not establish immutable model provenance for controlled production releases.
- Synchronous request limits cannot be solved merely by expanding the model context or retrying indefinitely.
- Container readiness checks alone do not guarantee successful predictions under real traffic and data distributions.
Explanation
Correct Answer
C:Synchronous request limits cannot be solved merely by expanding the model context or retrying indefinitely. This explains the mechanism behind online prediction timeout and batch fallback and why the described failure is possible. A discriminating investigation is: Review payload size distributions, request duration, response timeout policies and tolerance for deferred completion. The corresponding repair is: Use asynchronous or batch processing with durable job status for oversized workloads, reserving online calls for bounded requests.
Incorrect Answers
A:Network distance can dominate inference response times when the actual model execution is already short. This principle describes serving region proximity versus model compute, which is not the cause of the observed online prediction timeout and batch fallback failure. The controlling explanation is: Synchronous request limits cannot be solved merely by expanding the model context or retrying indefinitely.
B:Mutable nicknames do not establish immutable model provenance for controlled production releases. This principle describes model registry version provenance, which is not the cause of the observed online prediction timeout and batch fallback failure. The controlling explanation is: Synchronous request limits cannot be solved merely by expanding the model context or retrying indefinitely.
D:Container readiness checks alone do not guarantee successful predictions under real traffic and data distributions. This principle describes model traffic rollback on high error rate, which is not the cause of the observed online prediction timeout and batch fallback failure. The controlling explanation is: Synchronous request limits cannot be solved merely by expanding the model context or retrying indefinitely.
Question 67
During a regulated MLOps pipeline deployment, the following condition occurs: A distributed batch preprocessing workflow launches thousands of Ray tasks but loses intermediate outputs after one worker fails because object references are not persisted appropriately. The workflow must be repeatable and protect the last approved production model. Which outcome most directly demonstrates a successful, measurable fix?
- Small harmless fluctuations do not force releases and genuine regressions invoke a tested rollback or controlled retraining.
- A deliberately vulnerable test image is blocked from production while clean images pass the same gate.
- Changing the dataset version forces a rebuild while identical inputs may reuse validated cached results.
- A controlled worker loss resumes from saved progress and produces consistent output counts.
Explanation
Correct Answer
D:A controlled worker loss resumes from saved progress and produces consistent output counts. This is specific acceptance evidence for ray distributed task failure isolation. Distributed in-memory object stores are not durable archival storage; failure recovery needs explicit design. It confirms the controlled correction: Use supported Ray retry policies and durable checkpoints at suitable stage boundaries instead of relying only on worker memory.
Incorrect Answers
A:Small harmless fluctuations do not force releases and genuine regressions invoke a tested rollback or controlled retraining. This result could demonstrate success for canary failure feedback into ct policy but not for the stated ray distributed task failure isolation issue. The direct test is: A controlled worker loss resumes from saved progress and produces consistent output counts.
B:A deliberately vulnerable test image is blocked from production while clean images pass the same gate. This result could demonstrate success for cloud build container security gate but not for the stated ray distributed task failure isolation issue. The direct test is: A controlled worker loss resumes from saved progress and produces consistent output counts.
C:Changing the dataset version forces a rebuild while identical inputs may reuse validated cached results. This result could demonstrate success for kubeflow component cache invalidation but not for the stated ray distributed task failure isolation issue. The direct test is: A controlled worker loss resumes from saved progress and produces consistent output counts.
Question 68
During a postdeployment AI safety and reliability review, the following condition occurs: A transaction fraud model's feature distributions stay nearly unchanged while the relationship between inputs and observed fraud outcomes deteriorates. The model owner must provide repeatable evidence rather than a cosmetic dashboard change. What acceptance result verifies the intended behavior after remediation?
- Precision and recall recover on new labeled periods and monitoring distinguishes performance change from mere input shift.
- An evaluation fixture correctly associates each matured label with exactly one prior prediction.
- The assistant meets groundedness and task quality thresholds with reduced tokens and p95 latency.
- Attack-bearing documents fail to redirect privileged behavior while legitimate grounded answers remain available.
Explanation
Correct Answer
A:Precision and recall recover on new labeled periods and monitoring distinguishes performance change from mere input shift. This is specific acceptance evidence for concept drift with stable feature statistics. Concept drift changes conditional label relationships and may occur without obvious marginal feature drift. It confirms the controlled correction: Monitor outcome-based quality metrics on matured cohorts and retrain when verified concept drift is significant.
Incorrect Answers
B:An evaluation fixture correctly associates each matured label with exactly one prior prediction. This result could demonstrate success for ground truth association for quality monitoring but not for the stated concept drift with stable feature statistics issue. The direct test is: Precision and recall recover on new labeled periods and monitoring distinguishes performance change from mere input shift.
C:The assistant meets groundedness and task quality thresholds with reduced tokens and p95 latency. This result could demonstrate success for llm latency and cost token regression but not for the stated concept drift with stable feature statistics issue. The direct test is: Precision and recall recover on new labeled periods and monitoring distinguishes performance change from mere input shift.
D:Attack-bearing documents fail to redirect privileged behavior while legitimate grounded answers remain available. This result could demonstrate success for prompt injection through retrieved documents but not for the stated concept drift with stable feature statistics issue. The direct test is: Precision and recall recover on new labeled periods and monitoring distinguishes performance change from mere input shift.
Question 69
At a Google Cloud analytics team preparing a governed pilot, the following condition occurs: A customer churn dataset has yes/no labels but an analyst trains unsupervised k-means clusters and treats cluster IDs as churn probabilities. The team must satisfy a reproducible low-code AI assessment. Which TWO selections represent the correct investigation and remediation together? Select TWO.
- Train a supervised BigQuery ML LOGISTIC_REG or supported classifier with the churn label and evaluate precision-recall behavior.
- Inspect feature timestamps, label generation time, feature attributions, and whether the evaluation partition is independent.
- Move deterministic feature transformations into the supported BigQuery ML TRANSFORM workflow and use the same model at inference.
- Inspect the BOOLEAN or categorical outcome, cluster assignments, predicted probabilities, and the intended supervised evaluation metric.
- Benchmark a supported lower-latency Gemini variant against the current model under representative traffic before routing requests.
Explanation
Correct Answers
A:Train a supervised BigQuery ML LOGISTIC_REG or supported classifier with the churn label and evaluate precision-recall behavior. This is the targeted corrective response for bigquery ml logistic model for conversion probability. Clustering segments observations without learning their known outcomes, while supervised classification predicts those outcomes. Verify the diagnosis using: Inspect the BOOLEAN or categorical outcome, cluster assignments, predicted probabilities, and the intended supervised evaluation metric. Confirm success using: An unseen labeled cohort receives probabilities for the intended positive class with a calibrated operating threshold.
D:Inspect the BOOLEAN or categorical outcome, cluster assignments, predicted probabilities, and the intended supervised evaluation metric. This inspects evidence relevant to bigquery ml logistic model for conversion probability. Clustering segments observations without learning their known outcomes, while supervised classification predicts those outcomes. After identifying the cause, the matching correction is: Train a supervised BigQuery ML LOGISTIC_REG or supported classifier with the churn label and evaluate precision-recall behavior.
Incorrect Answers
B:Inspect feature timestamps, label generation time, feature attributions, and whether the evaluation partition is independent. This is an action or diagnostic relevant to bigquery ml ml.evaluate leakage detection, not one of the required two steps. In this scenario verify: Inspect the BOOLEAN or categorical outcome, cluster assignments, predicted probabilities, and the intended supervised evaluation metric. Then correct: Train a supervised BigQuery ML LOGISTIC_REG or supported classifier with the churn label and evaluate precision-recall behavior.
C:Move deterministic feature transformations into the supported BigQuery ML TRANSFORM workflow and use the same model at inference. This is an action or diagnostic relevant to bigquery ml transform serving parity, not one of the required two steps. In this scenario verify: Inspect the BOOLEAN or categorical outcome, cluster assignments, predicted probabilities, and the intended supervised evaluation metric. Then correct: Train a supervised BigQuery ML LOGISTIC_REG or supported classifier with the churn label and evaluate precision-recall behavior.
E:Benchmark a supported lower-latency Gemini variant against the current model under representative traffic before routing requests. This is an action or diagnostic relevant to gemini model choice for latency versus complexity, not one of the required two steps. In this scenario verify: Inspect the BOOLEAN or categorical outcome, cluster assignments, predicted probabilities, and the intended supervised evaluation metric. Then correct: Train a supervised BigQuery ML LOGISTIC_REG or supported classifier with the churn label and evaluate precision-recall behavior.
Question 70
While preparing a production training workflow on Gemini Enterprise Agent Platform, the following condition occurs: An eight-GPU job scales only slightly beyond two GPUs because each small batch requires excessive cross-worker gradient synchronization. Engineers need an auditable, resource-aware correction to the issue. Which evidence most directly establishes what is wrong with this AI or ML workflow?
- Inspect region-zone capacity, quota limits, accelerator requirements and scheduling error events.
- Inspect required output type, business loss, sample labels, and the configured model objective and metric.
- Inspect model framework, accelerator kernel support, batch sizes, input pipeline speed and per-step cost.
- Inspect all-reduce time, network bandwidth, effective batch size, worker stragglers and samples per second.
Explanation
Correct Answer
D:Inspect all-reduce time, network bandwidth, effective batch size, worker stragglers and samples per second. This inspects evidence relevant to distributed gradient communication bottleneck. Distributed speedups depend on communication-to-compute ratio, not merely the number of accelerators allocated. After identifying the cause, the matching correction is: Tune supported distributed batch sizing, communication strategy or placement after measuring the bottleneck.
Incorrect Answers
A:Inspect region-zone capacity, quota limits, accelerator requirements and scheduling error events. This examines accelerator quota and regional availability rather than the specific mechanism reported here. It is not the decisive first check for distributed gradient communication bottleneck. The relevant evidence is: Inspect all-reduce time, network bandwidth, effective batch size, worker stragglers and samples per second.
B:Inspect required output type, business loss, sample labels, and the configured model objective and metric. This examines regression versus classification product choice rather than the specific mechanism reported here. It is not the decisive first check for distributed gradient communication bottleneck. The relevant evidence is: Inspect all-reduce time, network bandwidth, effective batch size, worker stragglers and samples per second.
C:Inspect model framework, accelerator kernel support, batch sizes, input pipeline speed and per-step cost. This examines tpu versus gpu for tensorflow workload rather than the specific mechanism reported here. It is not the decisive first check for distributed gradient communication bottleneck. The relevant evidence is: Inspect all-reduce time, network bandwidth, effective batch size, worker stragglers and samples per second.