Exam Bundle

An unapproved model package is rejected and only a reviewed approved package reaches the endpoint update step. This would verify a different operation: Require the intended registered model package approval state and enforce a deployment role boundary at the promotion stage. That separate action addresses: A CodePipeline deployment stage accepts any model registry package regardless of its manual approval status. It cannot establish the required Silvercrest University correction for: A team serves a tiny low-traffic model on a provisioned GPU endpoint even though CPU inference meets the request latency target. The decisive acceptance check is instead: The representative load test meets latency targets and the chosen hosting cost falls below the idle GPU baseline.

Only an approved data manifest triggers retraining and repeated notifications do not launch duplicate experiments. This would verify a different operation: Filter EventBridge or S3-trigger inputs to the intended versioned training dataset and apply an idempotent retraining key. That separate action addresses: An automated model retraining workflow starts every time any S3 object changes, including log and metadata files unrelated to training. It cannot establish the required Silvercrest University correction for: A team serves a tiny low-traffic model on a provisioned GPU endpoint even though CPU inference meets the request latency target. The decisive acceptance check is instead: The representative load test meets latency targets and the chosen hosting cost falls below the idle GPU baseline.

Two identical trigger messages resolve to the expected single logical model update without competing untracked releases. This would verify a different operation: Include stable dataset IDs and idempotency controls in the orchestration design, preserving run lineage and concurrency policies. That separate action addresses: A retraining pipeline can be launched twice for the same input data, creating competing model versions with indistinguishable metadata. It cannot establish the required Silvercrest University correction for: A team serves a tiny low-traffic model on a provisioned GPU endpoint even though CPU inference meets the request latency target. The decisive acceptance check is instead: The representative load test meets latency targets and the chosen hosting cost falls below the idle GPU baseline.

Separate pipeline, build, training and serving roles with narrowly scoped permissions for their specific stages. It does not directly address the Harborview Insurance case: A production endpoint hosts hundreds of small customer-specific models on separate instances, creating high fixed infrastructure cost. The relevant acceptance criterion for this case is: Representative models load on demand and per-model latency and isolation checks meet the accepted limits at reduced aggregate cost.

Design the artifact sharing, replication and KMS policies explicitly for each trusted cross-account stage. It does not directly address the Harborview Insurance case: A production endpoint hosts hundreds of small customer-specific models on separate instances, creating high fixed infrastructure cost. The relevant acceptance criterion for this case is: Representative models load on demand and per-model latency and isolation checks meet the accepted limits at reduced aggregate cost.

Add post-deployment smoke tests and monitored acceptance gates before the production release is marked successful. It does not directly address the Harborview Insurance case: A production endpoint hosts hundreds of small customer-specific models on separate instances, creating high fixed infrastructure cost. The relevant acceptance criterion for this case is: Representative models load on demand and per-model latency and isolation checks meet the accepted limits at reduced aggregate cost.

An increasing request rate triggers timely scaling and p95 inference latency remains within the approved limit. This would verify a different operation: Select an appropriate SageMaker endpoint scaling metric such as invocations per instance and tune thresholds from measured traffic. That separate action addresses: A scalable inference endpoint uses CPU utilization as its only target metric despite requests waiting in invocation queues and little CPU usage. It cannot establish the required Harborview Insurance correction for: A production endpoint hosts hundreds of small customer-specific models on separate instances, creating high fixed infrastructure cost. The decisive acceptance check is instead: Representative models load on demand and per-model latency and isolation checks meet the accepted limits at reduced aggregate cost.

A replacement task starts from the approved artifact version without needing an untracked disk originally attached to another node. This would verify a different operation: Use the intended durable artifact storage pattern and reproducible infrastructure definitions rather than node-local attachment assumptions. That separate action addresses: A model deployment relies on one manually created EBS volume that a replacement task cannot attach after failure. It cannot establish the required Harborview Insurance correction for: A production endpoint hosts hundreds of small customer-specific models on separate instances, creating high fixed infrastructure cost. The decisive acceptance check is instead: Representative models load on demand and per-model latency and isolation checks meet the accepted limits at reduced aggregate cost.

An unapproved model package is rejected and only a reviewed approved package reaches the endpoint update step. This would verify a different operation: Require the intended registered model package approval state and enforce a deployment role boundary at the promotion stage. That separate action addresses: A CodePipeline deployment stage accepts any model registry package regardless of its manual approval status. It cannot establish the required Harborview Insurance correction for: A production endpoint hosts hundreds of small customer-specific models on separate instances, creating high fixed infrastructure cost. The decisive acceptance check is instead: Representative models load on demand and per-model latency and isolation checks meet the accepted limits at reduced aggregate cost.

Select an appropriate SageMaker endpoint scaling metric such as invocations per instance and tune thresholds from measured traffic. It does not directly address the Oakridge Public Services case: A production endpoint hosts hundreds of small customer-specific models on separate instances, creating high fixed infrastructure cost. The relevant acceptance criterion for this case is: Representative models load on demand and per-model latency and isolation checks meet the accepted limits at reduced aggregate cost.

Version the task definition, container image and service deployment configuration as reviewed infrastructure changes. It does not directly address the Oakridge Public Services case: A production endpoint hosts hundreds of small customer-specific models on separate instances, creating high fixed infrastructure cost. The relevant acceptance criterion for this case is: Representative models load on demand and per-model latency and isolation checks meet the accepted limits at reduced aggregate cost.

Provision the approved VPC endpoint or controlled network route and its route table and access policies for model-artifact reads. It does not directly address the Oakridge Public Services case: A production endpoint hosts hundreds of small customer-specific models on separate instances, creating high fixed infrastructure cost. The relevant acceptance criterion for this case is: Representative models load on demand and per-model latency and isolation checks meet the accepted limits at reduced aggregate cost.

A destination test role retrieves the artifact and decrypts it while an unrelated account remains denied. This would verify a different operation: Design the artifact sharing, replication and KMS policies explicitly for each trusted cross-account stage. That separate action addresses: A cross-account promotion pipeline copies model artifacts but loses the KMS permissions required for the destination account to load them. It cannot establish the required Oakridge Public Services correction for: A production endpoint hosts hundreds of small customer-specific models on separate instances, creating high fixed infrastructure cost. The decisive acceptance check is instead: Representative models load on demand and per-model latency and isolation checks meet the accepted limits at reduced aggregate cost.

An unapproved model package is rejected and only a reviewed approved package reaches the endpoint update step. This would verify a different operation: Require the intended registered model package approval state and enforce a deployment role boundary at the promotion stage. That separate action addresses: A CodePipeline deployment stage accepts any model registry package regardless of its manual approval status. It cannot establish the required Oakridge Public Services correction for: A production endpoint hosts hundreds of small customer-specific models on separate instances, creating high fixed infrastructure cost. The decisive acceptance check is instead: Representative models load on demand and per-model latency and isolation checks meet the accepted limits at reduced aggregate cost.

An increasing request rate triggers timely scaling and p95 inference latency remains within the approved limit. This would verify a different operation: Select an appropriate SageMaker endpoint scaling metric such as invocations per instance and tune thresholds from measured traffic. That separate action addresses: A scalable inference endpoint uses CPU utilization as its only target metric despite requests waiting in invocation queues and little CPU usage. It cannot establish the required Oakridge Public Services correction for: A production endpoint hosts hundreds of small customer-specific models on separate instances, creating high fixed infrastructure cost. The decisive acceptance check is instead: Representative models load on demand and per-model latency and isolation checks meet the accepted limits at reduced aggregate cost.

Tune target tracking, cooldown and minimum capacity from observed workload periodicity rather than setting a zero-tolerance threshold. It does not directly address the Meadowbrook Finance case: An inference service chains preprocessing and prediction in application code that cannot be versioned with the SageMaker deployed model. The relevant acceptance criterion for this case is: A known request traverses the versioned preprocessing and prediction stages and produces the expected output after deployment.

Include explicitly approved per-segment acceptance checks and fairness review in the conditional deployment gates. It does not directly address the Meadowbrook Finance case: An inference service chains preprocessing and prediction in application code that cannot be versioned with the SageMaker deployed model. The relevant acceptance criterion for this case is: A known request traverses the versioned preprocessing and prediction stages and produces the expected output after deployment.

Separate infrastructure deployment roles from runtime execution roles with narrowly scoped resource access policies. It does not directly address the Meadowbrook Finance case: An inference service chains preprocessing and prediction in application code that cannot be versioned with the SageMaker deployed model. The relevant acceptance criterion for this case is: A known request traverses the versioned preprocessing and prediction stages and produces the expected output after deployment.

The deployed task reports the exact image digest emitted by the reviewed build stage. This would verify a different operation: Pass an immutable ECR image digest as a build artifact to the deployment stage rather than reconstructing a tag. That separate action addresses: A CodeBuild job pushes an inference image to ECR but the subsequent deploy stage uses a different mutable image tag. It cannot establish the required Meadowbrook Finance correction for: An inference service chains preprocessing and prediction in application code that cannot be versioned with the SageMaker deployed model. The decisive acceptance check is instead: A known request traverses the versioned preprocessing and prediction stages and produces the expected output after deployment.

A controlled deployment must pass endpoint health, representative inference and alarm checks to reach the completed state. This would verify a different operation: Add post-deployment smoke tests and monitored acceptance gates before the production release is marked successful. That separate action addresses: A team declares its pipeline finished when model training succeeds, though deployment smoke tests and runtime metrics are never observed. It cannot establish the required Meadowbrook Finance correction for: An inference service chains preprocessing and prediction in application code that cannot be versioned with the SageMaker deployed model. The decisive acceptance check is instead: A known request traverses the versioned preprocessing and prediction stages and produces the expected output after deployment.

A replay of transient spikes avoids oscillation while sustained load still triggers the required scale-out action. This would verify a different operation: Tune target tracking, cooldown and minimum capacity from observed workload periodicity rather than setting a zero-tolerance threshold. That separate action addresses: An auto-scaled inference endpoint scales out during every brief request spike and incurs unnecessary instance churn and warm-up overhead. It cannot establish the required Meadowbrook Finance correction for: An inference service chains preprocessing and prediction in application code that cannot be versioned with the SageMaker deployed model. The decisive acceptance check is instead: A known request traverses the versioned preprocessing and prediction stages and produces the expected output after deployment.

Define a lightweight readiness and liveness interface distinct from resource-intensive inference operations. It does not directly address the Redstone Aviation case: An inference service chains preprocessing and prediction in application code that cannot be versioned with the SageMaker deployed model. The relevant acceptance criterion for this case is: A known request traverses the versioned preprocessing and prediction stages and produces the expected output after deployment.

Pass an immutable ECR image digest as a build artifact to the deployment stage rather than reconstructing a tag. It does not directly address the Redstone Aviation case: An inference service chains preprocessing and prediction in application code that cannot be versioned with the SageMaker deployed model. The relevant acceptance criterion for this case is: A known request traverses the versioned preprocessing and prediction stages and produces the expected output after deployment.

Provision test environments from IaC and implement a reviewed cleanup action with protective rules for shared resources. It does not directly address the Redstone Aviation case: An inference service chains preprocessing and prediction in application code that cannot be versioned with the SageMaker deployed model. The relevant acceptance criterion for this case is: A known request traverses the versioned preprocessing and prediction stages and produces the expected output after deployment.

A deliberately underperforming candidate remains unapproved and a passing candidate records its evaluation evidence. This would verify a different operation: Add evaluation and conditional model-quality steps linked to the model registry approval process. That separate action addresses: A SageMaker Pipelines workflow trains and deploys without recording evaluation output, so every candidate is automatically approved. It cannot establish the required Redstone Aviation correction for: An inference service chains preprocessing and prediction in application code that cannot be versioned with the SageMaker deployed model. The decisive acceptance check is instead: A known request traverses the versioned preprocessing and prediction stages and produces the expected output after deployment.

A simulated bad candidate triggers the rollback and the restored endpoint returns the documented prior model responses. This would verify a different operation: Implement a controlled canary or blue-green rollout with CloudWatch alarm gates and automatic rollback to the approved prior version. That separate action addresses: A model's automatic deployment increases error rates, but the pipeline has no health-driven rollback condition or retained previous version. It cannot establish the required Redstone Aviation correction for: An inference service chains preprocessing and prediction in application code that cannot be versioned with the SageMaker deployed model. The decisive acceptance check is instead: A known request traverses the versioned preprocessing and prediction stages and produces the expected output after deployment.

An auditor resolves a production model to the exact source commit that passed the original automated tests. This would verify a different operation: Pin reviewed source commits or release tags and record their identities with immutable model and image artifacts. That separate action addresses: A model endpoint is redeployed from a branch after its source is rebased, making the exact source revision impossible to reconstruct. It cannot establish the required Redstone Aviation correction for: An inference service chains preprocessing and prediction in application code that cannot be versioned with the SageMaker deployed model. The decisive acceptance check is instead: A known request traverses the versioned preprocessing and prediction stages and produces the expected output after deployment.

Filter EventBridge or S3-trigger inputs to the intended versioned training dataset and apply an idempotent retraining key. It does not directly address the Highland Education case: An edge camera must infer while offline, but the model architecture depends on a continuously reachable cloud real-time endpoint. The relevant acceptance criterion for this case is: The disconnected edge device completes an approved inference and later reports its model version when connectivity returns.

Include explicitly approved per-segment acceptance checks and fairness review in the conditional deployment gates. It does not directly address the Highland Education case: An edge camera must infer while offline, but the model architecture depends on a continuously reachable cloud real-time endpoint. The relevant acceptance criterion for this case is: The disconnected edge device completes an approved inference and later reports its model version when connectivity returns.

Version the task definition, container image and service deployment configuration as reviewed infrastructure changes. It does not directly address the Highland Education case: An edge camera must infer while offline, but the model architecture depends on a continuously reachable cloud real-time endpoint. The relevant acceptance criterion for this case is: The disconnected edge device completes an approved inference and later reports its model version when connectivity returns.

A replacement task starts from the approved artifact version without needing an untracked disk originally attached to another node. This would verify a different operation: Use the intended durable artifact storage pattern and reproducible infrastructure definitions rather than node-local attachment assumptions. That separate action addresses: A model deployment relies on one manually created EBS volume that a replacement task cannot attach after failure. It cannot establish the required Highland Education correction for: An edge camera must infer while offline, but the model architecture depends on a continuously reachable cloud real-time endpoint. The decisive acceptance check is instead: The disconnected edge device completes an approved inference and later reports its model version when connectivity returns.

A replay of the execution reads identical object versions and generates a traceable, comparable training run. This would verify a different operation: Make the retraining input a named immutable dataset manifest and persist it in the pipeline execution parameters. That separate action addresses: A scheduled retraining pipeline uses the latest available dataset rather than the dataset cut recorded in its review ticket. It cannot establish the required Highland Education correction for: An edge camera must infer while offline, but the model architecture depends on a continuously reachable cloud real-time endpoint. The decisive acceptance check is instead: The disconnected edge device completes an approved inference and later reports its model version when connectivity returns.

A simulated bad candidate triggers the rollback and the restored endpoint returns the documented prior model responses. This would verify a different operation: Implement a controlled canary or blue-green rollout with CloudWatch alarm gates and automatic rollback to the approved prior version. That separate action addresses: A model's automatic deployment increases error rates, but the pipeline has no health-driven rollback condition or retained previous version. It cannot establish the required Highland Education correction for: An edge camera must infer while offline, but the model architecture depends on a continuously reachable cloud real-time endpoint. The decisive acceptance check is instead: The disconnected edge device completes an approved inference and later reports its model version when connectivity returns.