Exam Bundle

A controlled personal-data request produces timing and error records without the prohibited raw identifiers. This would verify a different operation: Redact or omit sensitive input fields from logs and emit privacy-safe structured metadata with governed retention. That separate action addresses: An inference endpoint logs complete customer request bodies including personal information even though troubleshooting needs only aggregate timings. It cannot establish the required Eastshore Energy correction for: A CloudWatch dashboard claims the inference host is idle because CPU is low, yet GPU memory saturation causes inference throttling. The decisive acceptance check is instead: The benchmark identifies the accelerator bottleneck and the selected corrective capacity change reduces inference failures.

A controlled malformed-input batch is counted in the rejection metric and overall error reporting includes all attempted records. This would verify a different operation: Track total received records, rejected records and successful predictions with clear denominators across batch processing. That separate action addresses: A batch prediction system drops malformed records but the monitoring report tracks only successfully scored inputs and understates total error rate. It cannot establish the required Eastshore Energy correction for: A CloudWatch dashboard claims the inference host is idle because CPU is low, yet GPU memory saturation causes inference throttling. The decisive acceptance check is instead: The benchmark identifies the accelerator bottleneck and the selected corrective capacity change reduces inference failures.

A controlled prediction from the new variant appears in the correct monitoring dataset and the old variant is not mistaken for current. This would verify a different operation: Update monitored endpoint or variant configuration and baseline references as part of the versioned deployment process. That separate action addresses: A risk model is deployed with a new model version but its SageMaker Model Monitor schedule continues capturing only the old endpoint variant. It cannot establish the required Eastshore Energy correction for: A CloudWatch dashboard claims the inference host is idle because CPU is low, yet GPU memory saturation causes inference throttling. The decisive acceptance check is instead: The benchmark identifies the accelerator bottleneck and the selected corrective capacity change reduces inference failures.

Integrate ECR image scanning and an enforced reviewed vulnerability acceptance gate before deploying the immutable image digest. It does not directly address the Riverton Utilities case: A scaling endpoint's queue grows despite spare traffic capacity because the service hits its allowed instance scaling quota. The relevant acceptance criterion for this case is: A controlled scale-out reaches the revised allowed capacity and queue depth returns to its operating threshold.

Use appropriate SageMaker Clarify post-training bias measures and valid governed outcome data with documented fairness interpretation. It does not directly address the Riverton Utilities case: A scaling endpoint's queue grows despite spare traffic capacity because the service hits its allowed instance scaling quota. The relevant acceptance criterion for this case is: A controlled scale-out reaches the revised allowed capacity and queue depth returns to its operating threshold.

Use approved private subnets, security groups and VPC endpoints where supported to restrict unnecessary network egress. It does not directly address the Riverton Utilities case: A scaling endpoint's queue grows despite spare traffic capacity because the service hits its allowed instance scaling quota. The relevant acceptance criterion for this case is: A controlled scale-out reaches the revised allowed capacity and queue depth returns to its operating threshold.

The training role reads its authorized feature objects but a test access to an unrelated bucket is denied. This would verify a different operation: Limit the SageMaker execution role and bucket policy to the approved data prefix and required object operations. That separate action addresses: An ML training execution role has wildcard permission to read every company S3 bucket although it needs one versioned feature prefix. It cannot establish the required Riverton Utilities correction for: A scaling endpoint's queue grows despite spare traffic capacity because the service hits its allowed instance scaling quota. The decisive acceptance check is instead: A controlled scale-out reaches the revised allowed capacity and queue depth returns to its operating threshold.

A deliberately miscalibrated sample breaches the documented reliability threshold while a well-calibrated control remains acceptable. This would verify a different operation: Monitor predicted probability buckets against observed outcome frequencies using suitable delayed labels and calibration baselines. That separate action addresses: A quality dashboard plots only the mean model score, hiding gradual calibration drift as actual event rates change over time. It cannot establish the required Riverton Utilities correction for: A scaling endpoint's queue grows despite spare traffic capacity because the service hits its allowed instance scaling quota. The decisive acceptance check is instead: A controlled scale-out reaches the revised allowed capacity and queue depth returns to its operating threshold.

A simulated untrusted pull request cannot assume the production release role and an approved release still deploys successfully. This would verify a different operation: Enforce protected build contexts, isolated untrusted code execution and restricted deployment credentials available only after review. That separate action addresses: An automated ML CI/CD runner accepts source contributions from untrusted forks while holding direct privileges to production SageMaker resources. It cannot establish the required Riverton Utilities correction for: A scaling endpoint's queue grows despite spare traffic capacity because the service hits its allowed instance scaling quota. The decisive acceptance check is instead: A controlled scale-out reaches the revised allowed capacity and queue depth returns to its operating threshold.

Publish an approved data subset or governed view and grant read access only to the required machine learning project roles. It does not directly address the Cedarfield Hospitality case: A scaling endpoint's queue grows despite spare traffic capacity because the service hits its allowed instance scaling quota. The relevant acceptance criterion for this case is: A controlled scale-out reaches the revised allowed capacity and queue depth returns to its operating threshold.

Limit the SageMaker execution role and bucket policy to the approved data prefix and required object operations. It does not directly address the Cedarfield Hospitality case: A scaling endpoint's queue grows despite spare traffic capacity because the service hits its allowed instance scaling quota. The relevant acceptance criterion for this case is: A controlled scale-out reaches the revised allowed capacity and queue depth returns to its operating threshold.

Use AWS Secrets Manager or an approved secret provider with a scoped runtime role and suppress secret values from pipeline logs. It does not directly address the Cedarfield Hospitality case: A scaling endpoint's queue grows despite spare traffic capacity because the service hits its allowed instance scaling quota. The relevant acceptance criterion for this case is: A controlled scale-out reaches the revised allowed capacity and queue depth returns to its operating threshold.

An authorized monitoring sample supports drift analysis and a security scan finds no forbidden raw identifiers in captured payloads. This would verify a different operation: Configure bounded and redacted inference data capture with access controls, sampling and retention matched to monitoring requirements. That separate action addresses: A SageMaker endpoint stores all prediction payloads in logs despite a governance rule prohibiting raw customer identifiers in captured inference data. It cannot establish the required Cedarfield Hospitality correction for: A scaling endpoint's queue grows despite spare traffic capacity because the service hits its allowed instance scaling quota. The decisive acceptance check is instead: A controlled scale-out reaches the revised allowed capacity and queue depth returns to its operating threshold.

A controlled malformed-input batch is counted in the rejection metric and overall error reporting includes all attempted records. This would verify a different operation: Track total received records, rejected records and successful predictions with clear denominators across batch processing. That separate action addresses: A batch prediction system drops malformed records but the monitoring report tracks only successfully scored inputs and understates total error rate. It cannot establish the required Cedarfield Hospitality correction for: A scaling endpoint's queue grows despite spare traffic capacity because the service hits its allowed instance scaling quota. The decisive acceptance check is instead: A controlled scale-out reaches the revised allowed capacity and queue depth returns to its operating threshold.

The development user can create approved experiments in the correct domain but receives AccessDenied for protected production operations. This would verify a different operation: Apply identity-based domain and resource permission boundaries with segregated workspace roles and least privilege. That separate action addresses: A SageMaker Studio user can create experiments in a production domain even though their intended access is restricted to development. It cannot establish the required Cedarfield Hospitality correction for: A scaling endpoint's queue grows despite spare traffic capacity because the service hits its allowed instance scaling quota. The decisive acceptance check is instead: A controlled scale-out reaches the revised allowed capacity and queue depth returns to its operating threshold.

Configure precise cross-account package, S3 and KMS key permissions for the trusted production consumer role. It does not directly address the Alpine Healthcare case: An ML API has intermittent latency spikes but CloudWatch cannot reveal whether the delay occurs in preprocessing or downstream inference. The relevant acceptance criterion for this case is: A representative slow request shows the delay attributed to one documented processing stage rather than undifferentiated total time.

Distinguish covariate drift from verified performance degradation and evaluate outcome labels and model-quality metrics separately. It does not directly address the Alpine Healthcare case: An ML API has intermittent latency spikes but CloudWatch cannot reveal whether the delay occurs in preprocessing or downstream inference. The relevant acceptance criterion for this case is: A representative slow request shows the delay attributed to one documented processing stage rather than undifferentiated total time.

Update monitored endpoint or variant configuration and baseline references as part of the versioned deployment process. It does not directly address the Alpine Healthcare case: An ML API has intermittent latency spikes but CloudWatch cannot reveal whether the delay occurs in preprocessing or downstream inference. The relevant acceptance criterion for this case is: A representative slow request shows the delay attributed to one documented processing stage rather than undifferentiated total time.

A reference set with known labels produces a measurable recall trend and an agreed quality alert fires when the threshold is breached. This would verify a different operation: Create a governed delayed-label collection path and compare production predictions with matched observed labels for model quality monitoring. That separate action addresses: A production classifier's false-negative rate rises over several months, but predictions are logged without collecting any eventual ground-truth outcomes. It cannot establish the required Alpine Healthcare correction for: An ML API has intermittent latency spikes but CloudWatch cannot reveal whether the delay occurs in preprocessing or downstream inference. The decisive acceptance check is instead: A representative slow request shows the delay attributed to one documented processing stage rather than undifferentiated total time.

New output artifacts have the documented KMS key metadata and an attempted unencrypted write is rejected by bucket policy. This would verify a different operation: Require appropriate SSE-KMS settings on the output bucket and ensure the training role is authorized to use the chosen key. That separate action addresses: A model training job writes regulated outputs to an unencrypted S3 location despite a documented data-at-rest encryption standard. It cannot establish the required Alpine Healthcare correction for: An ML API has intermittent latency spikes but CloudWatch cannot reveal whether the delay occurs in preprocessing or downstream inference. The decisive acceptance check is instead: A representative slow request shows the delay attributed to one documented processing stage rather than undifferentiated total time.

A tampered artifact is rejected before the endpoint loads it and an approved version passes integrity verification. This would verify a different operation: Pin a governed immutable model artifact version and validate checksums or signatures through the approved release process. That separate action addresses: A production model file is overwritten with a maliciously altered artifact because deployment does not verify integrity or origin. It cannot establish the required Alpine Healthcare correction for: An ML API has intermittent latency spikes but CloudWatch cannot reveal whether the delay occurs in preprocessing or downstream inference. The decisive acceptance check is instead: A representative slow request shows the delay attributed to one documented processing stage rather than undifferentiated total time.

Apply identity-based domain and resource permission boundaries with segregated workspace roles and least privilege. It does not directly address the Juniper Media case: An ML API has intermittent latency spikes but CloudWatch cannot reveal whether the delay occurs in preprocessing or downstream inference. The relevant acceptance criterion for this case is: A representative slow request shows the delay attributed to one documented processing stage rather than undifferentiated total time.

Publish an approved data subset or governed view and grant read access only to the required machine learning project roles. It does not directly address the Juniper Media case: An ML API has intermittent latency spikes but CloudWatch cannot reveal whether the delay occurs in preprocessing or downstream inference. The relevant acceptance criterion for this case is: A representative slow request shows the delay attributed to one documented processing stage rather than undifferentiated total time.

Run shadow or controlled A/B inference comparisons with logged model version and matched request IDs before full rollout. It does not directly address the Juniper Media case: An ML API has intermittent latency spikes but CloudWatch cannot reveal whether the delay occurs in preprocessing or downstream inference. The relevant acceptance criterion for this case is: A representative slow request shows the delay attributed to one documented processing stage rather than undifferentiated total time.

An injected malformed feature vector is handled according to policy and creates an attributable monitoring event. This would verify a different operation: Apply input validation with monitored missingness and a documented safe fallback or rejection policy for out-of-contract requests. That separate action addresses: A model receives rare null-heavy inputs in production that were absent from training and a batch of such requests produces unstable predictions. It cannot establish the required Juniper Media correction for: An ML API has intermittent latency spikes but CloudWatch cannot reveal whether the delay occurs in preprocessing or downstream inference. The decisive acceptance check is instead: A representative slow request shows the delay attributed to one documented processing stage rather than undifferentiated total time.

A deliberately vulnerable test image is blocked and a compliant approved digest reaches the deployment stage. This would verify a different operation: Integrate ECR image scanning and an enforced reviewed vulnerability acceptance gate before deploying the immutable image digest. That separate action addresses: A training pipeline publishes an unscanned container image with known vulnerable dependencies directly to the production ECR repository. It cannot establish the required Juniper Media correction for: An ML API has intermittent latency spikes but CloudWatch cannot reveal whether the delay occurs in preprocessing or downstream inference. The decisive acceptance check is instead: A representative slow request shows the delay attributed to one documented processing stage rather than undifferentiated total time.

A normal seasonal replay stays below the reviewed alert threshold and an injected abnormal shift still produces an alert. This would verify a different operation: Define seasonal reference windows and calibrated drift thresholds rather than using an inappropriate static training baseline for all periods. That separate action addresses: A model monitoring job alerts on every small weekly seasonal variation and overwhelms operators despite meeting business quality requirements. It cannot establish the required Juniper Media correction for: An ML API has intermittent latency spikes but CloudWatch cannot reveal whether the delay occurs in preprocessing or downstream inference. The decisive acceptance check is instead: A representative slow request shows the delay attributed to one documented processing stage rather than undifferentiated total time.

Create monitored business slices and evaluate approved per-segment metrics and bias indicators using retained ground-truth labels. It does not directly address the Timberline Foods case: A training environment repeatedly provisions expensive instances after data preparation finishes and leaves them idle until an operator manually stops them. The relevant acceptance criterion for this case is: A finished training execution releases its compute automatically and leaves only approved persistent artifacts for later evaluation.

Version monitoring baselines with each released model and capture a reviewed transition policy between old and new expectations. It does not directly address the Timberline Foods case: A training environment repeatedly provisions expensive instances after data preparation finishes and leaves them idle until an operator manually stops them. The relevant acceptance criterion for this case is: A finished training execution releases its compute automatically and leaves only approved persistent artifacts for later evaluation.

Use appropriate SageMaker Clarify post-training bias measures and valid governed outcome data with documented fairness interpretation. It does not directly address the Timberline Foods case: A training environment repeatedly provisions expensive instances after data preparation finishes and leaves them idle until an operator manually stops them. The relevant acceptance criterion for this case is: A finished training execution releases its compute automatically and leaves only approved persistent artifacts for later evaluation.

A controlled endpoint configuration change yields an attributable CloudTrail event retained at the authorized destination. This would verify a different operation: Enable appropriate organization or account CloudTrail trails with secure retention and access controls for relevant management events. That separate action addresses: A security auditor cannot trace who modified a SageMaker endpoint because AWS CloudTrail event retention and organization trails were never configured. It cannot establish the required Timberline Foods correction for: A training environment repeatedly provisions expensive instances after data preparation finishes and leaves them idle until an operator manually stops them. The decisive acceptance check is instead: A finished training execution releases its compute automatically and leaves only approved persistent artifacts for later evaluation.

A controlled malformed-input batch is counted in the rejection metric and overall error reporting includes all attempted records. This would verify a different operation: Track total received records, rejected records and successful predictions with clear denominators across batch processing. That separate action addresses: A batch prediction system drops malformed records but the monitoring report tracks only successfully scored inputs and understates total error rate. It cannot establish the required Timberline Foods correction for: A training environment repeatedly provisions expensive instances after data preparation finishes and leaves them idle until an operator manually stops them. The decisive acceptance check is instead: A finished training execution releases its compute automatically and leaves only approved persistent artifacts for later evaluation.

The monitoring metadata links the live model package to its correct reviewed baseline and tests produce expected alerts. This would verify a different operation: Version monitoring baselines with each released model and capture a reviewed transition policy between old and new expectations. That separate action addresses: A model owner uses one unversioned monitoring baseline for every retraining run, so legitimate model improvements trigger constant false alarms. It cannot establish the required Timberline Foods correction for: A training environment repeatedly provisions expensive instances after data preparation finishes and leaves them idle until an operator manually stops them. The decisive acceptance check is instead: A finished training execution releases its compute automatically and leaves only approved persistent artifacts for later evaluation.