Exam Bundle

The duplicate audit shows coherent labels within matching-image clusters and a reviewed exception log tracks ambiguous cases. This would verify a different operation: Fingerprint and review duplicate image groups and reconcile their class labels before publishing the approved training set. That separate action addresses: A classification training set contains duplicated near-identical product images assigned conflicting labels, creating avoidable noise. It cannot establish the required Bluehaven Transit correction for: The SageMaker Feature Store offline store is configured with a wrong S3 bucket and the expected historical training features never appear for query. The decisive acceptance check is instead: New feature records are queryable from the intended offline store after normal delivery delay while the online store remains functional.

A cross-split identity audit finds no patient overlap and the held-out metric is computed on previously unseen identities. This would verify a different operation: Split examples by patient identity and keep all records for one identity in only one dataset partition. That separate action addresses: An ML pipeline randomly splits multiple events from one patient across training and test datasets, leaking individual-specific signals. It cannot establish the required Bluehaven Transit correction for: The SageMaker Feature Store offline store is configured with a wrong S3 bucket and the expected historical training features never appear for query. The decisive acceptance check is instead: New feature records are queryable from the intended offline store after normal delivery delay while the online store remains functional.

The large synthetic partition completes without driver out-of-memory errors and its aggregate counts match the reference run. This would verify a different operation: Use distributed Spark transforms or scalable partitioned outputs instead of collecting the full dataset to one process. That separate action addresses: A Spark feature preparation stage converts all encoded records to driver memory and fails for large training partitions. It cannot establish the required Bluehaven Transit correction for: The SageMaker Feature Store offline store is configured with a wrong S3 bucket and the expected historical training features never appear for query. The decisive acceptance check is instead: New feature records are queryable from the intended offline store after normal delivery delay while the online store remains functional.

Fit and persist a categorical encoder on the training set with an explicit unknown-category handling policy for all later inputs. It does not directly address the Brookside Manufacturing case: The SageMaker Feature Store offline store is configured with a wrong S3 bucket and the expected historical training features never appear for query. The relevant acceptance criterion for this case is: New feature records are queryable from the intended offline store after normal delivery delay while the online store remains functional.

Run the appropriate Clarify data-bias analysis and record the fairness baseline and approved mitigation assessment before model release. It does not directly address the Brookside Manufacturing case: The SageMaker Feature Store offline store is configured with a wrong S3 bucket and the expected historical training features never appear for query. The relevant acceptance criterion for this case is: New feature records are queryable from the intended offline store after normal delivery delay while the online store remains functional.

Record and deploy a specific reviewed recipe version together with input snapshot identifiers for every training dataset build. It does not directly address the Brookside Manufacturing case: The SageMaker Feature Store offline store is configured with a wrong S3 bucket and the expected historical training features never appear for query. The relevant acceptance criterion for this case is: New feature records are queryable from the intended offline store after normal delivery delay while the online store remains functional.

An audit of transformed images preserves orientation-sensitive labels and prohibited flips are absent from model training. This would verify a different operation: Restrict image augmentation to clinically valid transformations and verify label invariance for each permitted operation. That separate action addresses: A clinical image dataset uses an augmentation pipeline that reverses left-right orientation despite labels depending on laterality. It cannot establish the required Brookside Manufacturing correction for: The SageMaker Feature Store offline store is configured with a wrong S3 bucket and the expected historical training features never appear for query. The decisive acceptance check is instead: New feature records are queryable from the intended offline store after normal delivery delay while the online store remains functional.

Every required rare class appears at least the configured minimum count in evaluation and no training row is duplicated into test. This would verify a different operation: Use a stratified or otherwise class-aware splitting method with declared minimum representation checks for the holdout set. That separate action addresses: A multi-class label dataset is split without stratification and a rare class is missing entirely from the evaluation partition. It cannot establish the required Brookside Manufacturing correction for: The SageMaker Feature Store offline store is configured with a wrong S3 bucket and the expected historical training features never appear for query. The decisive acceptance check is instead: New feature records are queryable from the intended offline store after normal delivery delay while the online store remains functional.

A repeated run references the same content hashes and rejects an unexpected overwrite of a previously accepted input object. This would verify a different operation: Enable controlled dataset versioning or publish immutable object checksums and a training input manifest for every model build. That separate action addresses: An S3 training dataset uses no immutable version identifier, so reruns silently train on objects updated during the model experiment. It cannot establish the required Brookside Manufacturing correction for: The SageMaker Feature Store offline store is configured with a wrong S3 bucket and the expected historical training features never appear for query. The decisive acceptance check is instead: New feature records are queryable from the intended offline store after normal delivery delay while the online store remains functional.

Define an as-of-time feature cutoff and exclude any data collected after the target prediction time. It does not directly address the Harborline Logistics case: A SageMaker processing job repeatedly downloads a multi-gigabyte model-input dataset from a distant public endpoint and fails on unstable external transfers. The relevant acceptance criterion for this case is: A rerun uses the same immutable S3 manifest and produces identical input checksums without public endpoint dependency.

Use chronological training, validation and final holdout windows with a gap where leakage risk requires one. It does not directly address the Harborline Logistics case: A SageMaker processing job repeatedly downloads a multi-gigabyte model-input dataset from a distant public endpoint and fails on unstable external transfers. The relevant acceptance criterion for this case is: A rerun uses the same immutable S3 manifest and produces identical input checksums without public endpoint dependency.

Add schema, null-rate and range checks using AWS Glue Data Quality or equivalent before the training step begins. It does not directly address the Harborline Logistics case: A SageMaker processing job repeatedly downloads a multi-gigabyte model-input dataset from a distant public endpoint and fails on unstable external transfers. The relevant acceptance criterion for this case is: A rerun uses the same immutable S3 manifest and produces identical input checksums without public endpoint dependency.

A training role cannot access the mapping store while an explicitly authorized re-identification role can complete an audited lookup. This would verify a different operation: Protect the token mapping with separate least-privilege access and encryption controls and limit its retention to approved workflows. That separate action addresses: An ML pipeline stores masked user IDs but uses a reversible mapping file in a publicly accessible S3 location alongside training data. It cannot establish the required Harborline Logistics correction for: A SageMaker processing job repeatedly downloads a multi-gigabyte model-input dataset from a distant public endpoint and fails on unstable external transfers. The decisive acceptance check is instead: A rerun uses the same immutable S3 manifest and produces identical input checksums without public endpoint dependency.

A bilingual reference set retains the required product terms and its label agreement remains inside the approved quality range. This would verify a different operation: Preserve protected product tokens or use a validated multilingual feature representation that does not corrupt business identifiers. That separate action addresses: A sentiment pipeline blindly translates all customer messages into English even though translation changes domain-specific product identifiers. It cannot establish the required Harborline Logistics correction for: A SageMaker processing job repeatedly downloads a multi-gigabyte model-input dataset from a distant public endpoint and fails on unstable external transfers. The decisive acceptance check is instead: A rerun uses the same immutable S3 manifest and produces identical input checksums without public endpoint dependency.

A known test image produces matching normalized pixel tensors in offline and online environments. This would verify a different operation: Standardize image dimensions, channel ordering and value normalization in the shared preprocessing artifact. That separate action addresses: An image classification pipeline resizes training images with one color-channel convention and production inputs with another. It cannot establish the required Harborline Logistics correction for: A SageMaker processing job repeatedly downloads a multi-gigabyte model-input dataset from a distant public endpoint and fails on unstable external transfers. The decisive acceptance check is instead: A rerun uses the same immutable S3 manifest and produces identical input checksums without public endpoint dependency.

Split the original records first, then apply synthetic oversampling exclusively to the training split. It does not directly address the Northgate Telecom case: A SageMaker processing job repeatedly downloads a multi-gigabyte model-input dataset from a distant public endpoint and fails on unstable external transfers. The relevant acceptance criterion for this case is: A rerun uses the same immutable S3 manifest and produces identical input checksums without public endpoint dependency.

Measure class balance and compare stratified resampling or class weighting on training data while preserving an untouched evaluation set. It does not directly address the Northgate Telecom case: A SageMaker processing job repeatedly downloads a multi-gigabyte model-input dataset from a distant public endpoint and fails on unstable external transfers. The relevant acceptance criterion for this case is: A rerun uses the same immutable S3 manifest and produces identical input checksums without public endpoint dependency.

Perform approved irreversible masking or tokenization before feature extraction and restrict the original identifier fields. It does not directly address the Northgate Telecom case: A SageMaker processing job repeatedly downloads a multi-gigabyte model-input dataset from a distant public endpoint and fails on unstable external transfers. The relevant acceptance criterion for this case is: A rerun uses the same immutable S3 manifest and produces identical input checksums without public endpoint dependency.

A generated training sample contains no prohibited payment fields and access to the original source is limited to authorized roles. This would verify a different operation: Classify and remove unnecessary protected columns before the ML staging layer and enforce restricted access to retained originals. That separate action addresses: A sensitive training corpus contains raw payment card fields that are not needed for the predictive task but are present in shared notebook extracts. It cannot establish the required Northgate Telecom correction for: A SageMaker processing job repeatedly downloads a multi-gigabyte model-input dataset from a distant public endpoint and fails on unstable external transfers. The decisive acceptance check is instead: A rerun uses the same immutable S3 manifest and produces identical input checksums without public endpoint dependency.

A historical example contains the category active at its original event timestamp instead of a later corrected label. This would verify a different operation: Use time-aware joins against effective-dated reference values to reconstruct the features available when each event occurred. That separate action addresses: A batch transformation joins slowly changing reference categories using today's labels for all historic examples, leaking future feature state. It cannot establish the required Northgate Telecom correction for: A SageMaker processing job repeatedly downloads a multi-gigabyte model-input dataset from a distant public endpoint and fails on unstable external transfers. The decisive acceptance check is instead: A rerun uses the same immutable S3 manifest and produces identical input checksums without public endpoint dependency.

The final report records positive-class recall and precision on the original distribution, not only inflated aggregate accuracy. This would verify a different operation: Measure class balance and compare stratified resampling or class weighting on training data while preserving an untouched evaluation set. That separate action addresses: A marketing propensity model has a class distribution of 98 percent negative examples and selection by accuracy alone hides poor positive detection. It cannot establish the required Northgate Telecom correction for: A SageMaker processing job repeatedly downloads a multi-gigabyte model-input dataset from a distant public endpoint and fails on unstable external transfers. The decisive acceptance check is instead: A rerun uses the same immutable S3 manifest and produces identical input checksums without public endpoint dependency.

Learn imputation statistics exclusively from the training split and reuse those fixed values on held-out datasets. It does not directly address the Pinecrest Hotels case: An ML data lake receives tiny telemetry objects every second and Spark processing time is dominated by listing and opening millions of files. The relevant acceptance criterion for this case is: A benchmark reads fewer files, preserves event totals and lowers processing overhead for the same approved date partition.

Enable controlled dataset versioning or publish immutable object checksums and a training input manifest for every model build. It does not directly address the Pinecrest Hotels case: An ML data lake receives tiny telemetry objects every second and Spark processing time is dominated by listing and opening millions of files. The relevant acceptance criterion for this case is: A benchmark reads fewer files, preserves event totals and lowers processing overhead for the same approved date partition.

Use chronological training, validation and final holdout windows with a gap where leakage risk requires one. It does not directly address the Pinecrest Hotels case: An ML data lake receives tiny telemetry objects every second and Spark processing time is dominated by listing and opening millions of files. The relevant acceptance criterion for this case is: A benchmark reads fewer files, preserves event totals and lowers processing overhead for the same approved date partition.

Model input checks confirm no implied numerical ordering and all expected regions map to the approved encoded features. This would verify a different operation: Apply an appropriate categorical encoding while preserving the region identifier's nominal semantics and missing-value policy. That separate action addresses: A numeric categorical code such as region ID is mistakenly treated as an ordered magnitude when supplied to a linear model. It cannot establish the required Pinecrest Hotels correction for: An ML data lake receives tiny telemetry objects every second and Spark processing time is dominated by listing and opening millions of files. The decisive acceptance check is instead: A benchmark reads fewer files, preserves event totals and lowers processing overhead for the same approved date partition.

The final report records positive-class recall and precision on the original distribution, not only inflated aggregate accuracy. This would verify a different operation: Measure class balance and compare stratified resampling or class weighting on training data while preserving an untouched evaluation set. That separate action addresses: A marketing propensity model has a class distribution of 98 percent negative examples and selection by accuracy alone hides poor positive detection. It cannot establish the required Pinecrest Hotels correction for: An ML data lake receives tiny telemetry objects every second and Spark processing time is dominated by listing and opening millions of files. The decisive acceptance check is instead: A benchmark reads fewer files, preserves event totals and lowers processing overhead for the same approved date partition.

An input set with mixed-capitalization labels maps consistently while an unknown label triggers the declared quality exception. This would verify a different operation: Define a validated label encoding dictionary and reject unmapped target values before model training starts. That separate action addresses: A binary target is stored as strings with inconsistent capitalization and preprocessing silently maps one valid positive class to negative. It cannot establish the required Pinecrest Hotels correction for: An ML data lake receives tiny telemetry objects every second and Spark processing time is dominated by listing and opening millions of files. The decisive acceptance check is instead: A benchmark reads fewer files, preserves event totals and lowers processing overhead for the same approved date partition.

Fingerprint and review duplicate image groups and reconcile their class labels before publishing the approved training set. It does not directly address the Silvercrest University case: An ML data lake receives tiny telemetry objects every second and Spark processing time is dominated by listing and opening millions of files. The relevant acceptance criterion for this case is: A benchmark reads fewer files, preserves event totals and lowers processing overhead for the same approved date partition.

Split the original records first, then apply synthetic oversampling exclusively to the training split. It does not directly address the Silvercrest University case: An ML data lake receives tiny telemetry objects every second and Spark processing time is dominated by listing and opening millions of files. The relevant acceptance criterion for this case is: A benchmark reads fewer files, preserves event totals and lowers processing overhead for the same approved date partition.

Choose a guarded nonnegative log transformation and test boundary handling before saving the feature pipeline. It does not directly address the Silvercrest University case: An ML data lake receives tiny telemetry objects every second and Spark processing time is dominated by listing and opening millions of files. The relevant acceptance criterion for this case is: A benchmark reads fewer files, preserves event totals and lowers processing overhead for the same approved date partition.

A test input with NULL is marked missing and a true zero-dollar record remains distinguishable in the clean dataset. This would verify a different operation: Validate nullable field semantics and reject or impute explicit missing tokens without converting them to legitimate zero values. That separate action addresses: A source CSV sends the literal string NULL for missing prices, but the preparation job treats it as a valid zero-dollar transaction. It cannot establish the required Silvercrest University correction for: An ML data lake receives tiny telemetry objects every second and Spark processing time is dominated by listing and opening millions of files. The decisive acceptance check is instead: A benchmark reads fewer files, preserves event totals and lowers processing overhead for the same approved date partition.

A cross-split identity audit finds no patient overlap and the held-out metric is computed on previously unseen identities. This would verify a different operation: Split examples by patient identity and keep all records for one identity in only one dataset partition. That separate action addresses: An ML pipeline randomly splits multiple events from one patient across training and test datasets, leaking individual-specific signals. It cannot establish the required Silvercrest University correction for: An ML data lake receives tiny telemetry objects every second and Spark processing time is dominated by listing and opening millions of files. The decisive acceptance check is instead: A benchmark reads fewer files, preserves event totals and lowers processing overhead for the same approved date partition.

A training role cannot access the mapping store while an explicitly authorized re-identification role can complete an audited lookup. This would verify a different operation: Protect the token mapping with separate least-privilege access and encryption controls and limit its retention to approved workflows. That separate action addresses: An ML pipeline stores masked user IDs but uses a reversible mapping file in a publicly accessible S3 location alongside training data. It cannot establish the required Silvercrest University correction for: An ML data lake receives tiny telemetry objects every second and Spark processing time is dominated by listing and opening millions of files. The decisive acceptance check is instead: A benchmark reads fewer files, preserves event totals and lowers processing overhead for the same approved date partition.