Google Professional Machine Learning Engineer Practice Test 79
Question 780
At a Google Cloud analytics team preparing a governed pilot, the following condition occurs: An analyst normalizes numerical inputs in a notebook before training but later calls ML.PREDICT on raw table columns, shifting prediction values. The team must satisfy a reproducible low-code AI assessment. What should the engineer inspect before selecting a correction?
- Inspect the target column type, business loss metric, and sample prediction schema alongside model_type in the BigQuery model metadata.
- Inspect the training preprocessing SQL, TRANSFORM clause, model input schema, and the inference query's feature values.
- Inspect representative invoices, desired structured fields, supported processor output and field-level extraction errors.
- Inspect label availability, feature types, tuning needs, governance requirements, and supported AutoML model interfaces.
Explanation
Correct Answer
B:Inspect the training preprocessing SQL, TRANSFORM clause, model input schema, and the inference query's feature values. This inspects evidence relevant to bigquery ml transform serving parity. Transformations embedded with a BigQuery ML model reduce divergence between fitting and serving features. After identifying the cause, the matching correction is: Move deterministic feature transformations into the supported BigQuery ML TRANSFORM workflow and use the same model at inference. Here, the relevant situation is: the following condition occurs: An analyst normalizes numerical inputs in a notebook before training but later calls….
Incorrect Answers
A:Inspect the target column type, business loss metric, and sample prediction schema alongside model_type in the BigQuery model metadata. This examines bigquery ml linear regression for continuous value rather than the specific mechanism reported here. It is not the decisive first check for bigquery ml transform serving parity. The relevant evidence is: Inspect the training preprocessing SQL, TRANSFORM clause, model input schema, and the inference query's feature values. Here, the relevant situation is: the following condition occurs: An analyst normalizes numerical inputs in a notebook before training but later calls….
C:Inspect representative invoices, desired structured fields, supported processor output and field-level extraction errors. This examines document ai structured invoice extraction rather than the specific mechanism reported here. It is not the decisive first check for bigquery ml transform serving parity. The relevant evidence is: Inspect the training preprocessing SQL, TRANSFORM clause, model input schema, and the inference query's feature values.
D:Inspect label availability, feature types, tuning needs, governance requirements, and supported AutoML model interfaces. This examines automl tabular supervised model selection rather than the specific mechanism reported here. It is not the decisive first check for bigquery ml transform serving parity. The relevant evidence is: Inspect the training preprocessing SQL, TRANSFORM clause, model input schema, and the inference query's feature values. Here, the relevant situation is: the following condition occurs: An analyst normalizes numerical inputs in a notebook before training but later calls….
Question 781
At a Google Cloud analytics team preparing a governed pilot, the following condition occurs: A team chooses the largest available foundation model by parameter count without checking whether its license allows commercial use or whether it handles the required input modality. The team must satisfy a reproducible low-code AI assessment. Which TWO complementary choices jointly establish the evidence and appropriate correction? Select TWO.
- Index versioned policy sources and use retrieval-grounded Gemini responses for changing facts, reserving tuning for justified behavior adaptation.
- Inspect requested labels or bounding boxes, sample image quality, supported Vision API capabilities and required detection thresholds.
- Use a suitable ARIMA_PLUS time-series model with timestamp and data columns and evaluate ML.FORECAST on rolling historical cutoffs.
- Review Model Garden model cards, permitted terms, input and output modalities, context limits and evaluation results on the target task.
- Select a permitted task-compatible Gemini or other supported foundation model and benchmark it against simpler candidates.
Explanation
Correct Answers
D:Review Model Garden model cards, permitted terms, input and output modalities, context limits and evaluation results on the target task. This inspects evidence relevant to model garden task and license selection. Model size alone does not ensure suitable task support, usage rights or production economics. After identifying the cause, the matching correction is: Select a permitted task-compatible Gemini or other supported foundation model and benchmark it against simpler candidates.
E:Select a permitted task-compatible Gemini or other supported foundation model and benchmark it against simpler candidates. This is the targeted corrective response for model garden task and license selection. Model size alone does not ensure suitable task support, usage rights or production economics. Verify the diagnosis using: Review Model Garden model cards, permitted terms, input and output modalities, context limits and evaluation results on the target task. Confirm success using: The chosen version meets task quality, inference cost and license rules on a documented representative evaluation set.
Incorrect Answers
A:Index versioned policy sources and use retrieval-grounded Gemini responses for changing facts, reserving tuning for justified behavior adaptation. This is an action or diagnostic relevant to foundation model tuning versus retrieval update, not one of the required two steps. In this scenario verify: Review Model Garden model cards, permitted terms, input and output modalities, context limits and evaluation results on the target task. Then correct: Select a permitted task-compatible Gemini or other supported foundation model and benchmark it against simpler candidates.
B:Inspect requested labels or bounding boxes, sample image quality, supported Vision API capabilities and required detection thresholds. This is an action or diagnostic relevant to vision api pretrained image labels versus custom detection, not one of the required two steps. In this scenario verify: Review Model Garden model cards, permitted terms, input and output modalities, context limits and evaluation results on the target task. Then correct: Select a permitted task-compatible Gemini or other supported foundation model and benchmark it against simpler candidates.
C:Use a suitable ARIMA_PLUS time-series model with timestamp and data columns and evaluate ML.FORECAST on rolling historical cutoffs. This is an action or diagnostic relevant to bigquery arima_plus forecast horizon, not one of the required two steps. In this scenario verify: Review Model Garden model cards, permitted terms, input and output modalities, context limits and evaluation results on the target task. Then correct: Select a permitted task-compatible Gemini or other supported foundation model and benchmark it against simpler candidates.
Question 782
During a postdeployment AI safety and reliability review, the following condition occurs: A product demand model raises a drift alert every weekend because the baseline ignores known weekly seasonality and monitoring thresholds are overly narrow. The model owner must provide repeatable evidence rather than a cosmetic dashboard change. What should the engineer inspect before selecting a correction?
- Inspect subgroup confusion matrices, sample sizes, training representation and evaluation slices over time.
- Inspect live feature histograms, training baselines, data-capture completeness and drift metric thresholds.
- Compare live and offline feature SQL, window definitions, sample entity values and prediction-time cutoffs.
- Examine alert history, day-of-week distributions, false-positive cost and baseline segmentation.
Explanation
Correct Answer
D:Examine alert history, day-of-week distributions, false-positive cost and baseline segmentation. This inspects evidence relevant to model monitoring false-alert seasonality. Normal recurring seasonality can resemble drift against an unsegmented static baseline. After identifying the cause, the matching correction is: Adopt justified seasonal baselines or cohorts and tune alert thresholds without suppressing severe shifts. This matters here because a product demand model raises a drift alert every weekend because the baseline ignores known weekly seasonality and….
Incorrect Answers
A:Inspect subgroup confusion matrices, sample sizes, training representation and evaluation slices over time. This examines model feature bias by subgroup rather than the specific mechanism reported here. It is not the decisive first check for model monitoring false-alert seasonality. The relevant evidence is: Examine alert history, day-of-week distributions, false-positive cost and baseline segmentation.
B:Inspect live feature histograms, training baselines, data-capture completeness and drift metric thresholds. This examines input data distribution drift alarm rather than the specific mechanism reported here. It is not the decisive first check for model monitoring false-alert seasonality. The relevant evidence is: Examine alert history, day-of-week distributions, false-positive cost and baseline segmentation.
C:Compare live and offline feature SQL, window definitions, sample entity values and prediction-time cutoffs. This examines training-serving feature skew detection rather than the specific mechanism reported here. It is not the decisive first check for model monitoring false-alert seasonality. The relevant evidence is: Examine alert history, day-of-week distributions, false-positive cost and baseline segmentation.
Question 783
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 targeted engineering change most directly addresses the failure?
- Publish one versioned pipeline package and promote it with reviewed environment-specific parameters.
- Require candidate evaluation against a governed baseline with an enforced acceptance threshold before promotion.
- Trigger candidate retraining only when justified evidence and sufficient labeled data are available, with evaluation gates.
- Use supported Ray retry policies and durable checkpoints at suitable stage boundaries instead of relying only on worker memory.
Explanation
Correct Answer
D:Use supported Ray retry policies and durable checkpoints at suitable stage boundaries instead of relying only on worker memory. This is the targeted corrective response for ray distributed task failure isolation. Distributed in-memory object stores are not durable archival storage; failure recovery needs explicit design. Verify the diagnosis using: Inspect task lineage, retries, durable checkpoints, object-store pressure and failed worker logs. Confirm success using: A controlled worker loss resumes from saved progress and produces consistent output counts. This matters here because a distributed batch preprocessing workflow launches thousands of Ray tasks but loses intermediate outputs after one….
Incorrect Answers
A:Publish one versioned pipeline package and promote it with reviewed environment-specific parameters. This action addresses pipeline environment parameter promotion, not the failure in the described ray distributed task failure isolation case. The targeted correction is: Use supported Ray retry policies and durable checkpoints at suitable stage boundaries instead of relying only on worker memory.
B:Require candidate evaluation against a governed baseline with an enforced acceptance threshold before promotion. This action addresses model evaluation gate before deployment, not the failure in the described ray distributed task failure isolation case. The targeted correction is: Use supported Ray retry policies and durable checkpoints at suitable stage boundaries instead of relying only on worker memory.
C:Trigger candidate retraining only when justified evidence and sufficient labeled data are available, with evaluation gates. This action addresses data drift triggered retraining policy, not the failure in the described ray distributed task failure isolation case. The targeted correction is: Use supported Ray retry policies and durable checkpoints at suitable stage boundaries instead of relying only on worker memory.
Question 784
During a controlled replay of the data and model handoff to a second engineering team, the following condition occurs: A distributed preprocessing job joins two huge feature tables, but one NULL-like customer category contains most rows and one reducer becomes the bottleneck. The handoff must preserve privacy, lineage and testable behavior. What acceptance result verifies the intended behavior after remediation?
- A clean notebook loads the exact model revision, runs representative inputs and reports consistent outputs.
- Long-tail shuffle tasks shorten and the normalized output row count matches a reference implementation.
- The scheduled query produces equivalent daily features with less service orchestration and within the run window.
- The notebook still trains on approved sample tables while restricted production datasets deny unauthorized reads.
Explanation
Correct Answer
B:Long-tail shuffle tasks shorten and the normalized output row count matches a reference implementation. This is specific acceptance evidence for managed spark join-key skew repair. Skew concentrates otherwise distributed work on a single hot key, which extra workers alone may not resolve. It confirms the controlled correction: Handle the dominant key separately or repartition and salt the skewed join while validating join semantics and counts.
Incorrect Answers
A:A clean notebook loads the exact model revision, runs representative inputs and reports consistent outputs. This result could demonstrate success for model garden jax accelerator compatibility but not for the stated managed spark join-key skew repair issue. The direct test is: Long-tail shuffle tasks shorten and the normalized output row count matches a reference implementation.
C:The scheduled query produces equivalent daily features with less service orchestration and within the run window. This result could demonstrate success for bigquery versus dataflow preparation selection but not for the stated managed spark join-key skew repair issue. The direct test is: Long-tail shuffle tasks shorten and the normalized output row count matches a reference implementation.
D:The notebook still trains on approved sample tables while restricted production datasets deny unauthorized reads. This result could demonstrate success for colab enterprise notebook data access scope but not for the stated managed spark join-key skew repair issue. The direct test is: Long-tail shuffle tasks shorten and the normalized output row count matches a reference implementation.
Question 785
During a controlled replay of the data and model handoff to a second engineering team, the following condition occurs: A support assistant team reports a model improvement based solely on an unrestricted judge model's one-run preference, ignoring reference criteria and judge bias. The handoff must preserve privacy, lineage and testable behavior. Which TWO complementary choices jointly establish the evidence and appropriate correction? Select TWO.
- Log reproducibility settings and rerun comparable trials with fixed split and controlled randomness before attributing changes.
- Build a constrained judge rubric with blind repeated comparisons and human-adjudicated calibration on representative examples.
- Redact or tokenize unnecessary identifying fields before labeling and limit access to approved mappings.
- Inspect Agent Platform ML Metadata artifacts, pipeline executions, dataset fingerprints and deployment revisions.
- Inspect judge rubric prompts, evaluation dataset balance, repeated ratings, human benchmark and uncertainty in pairwise preferences.
Explanation
Correct Answers
B:Build a constrained judge rubric with blind repeated comparisons and human-adjudicated calibration on representative examples. This is the targeted corrective response for llm-as-a-judge rubric reliability. LLM-based scoring can be sensitive to phrasing, order and model preferences; evaluation needs explicit calibration. Verify the diagnosis using: Inspect judge rubric prompts, evaluation dataset balance, repeated ratings, human benchmark and uncertainty in pairwise preferences. Confirm success using: Judge agreement against human standards improves and the decision is stable across repeated randomized evaluations. This matters here because during a controlled replay of the data and model handoff to a second engineering team, the following condition….
E:Inspect judge rubric prompts, evaluation dataset balance, repeated ratings, human benchmark and uncertainty in pairwise preferences. This inspects evidence relevant to llm-as-a-judge rubric reliability. LLM-based scoring can be sensitive to phrasing, order and model preferences; evaluation needs explicit calibration. After identifying the cause, the matching correction is: Build a constrained judge rubric with blind repeated comparisons and human-adjudicated calibration on representative examples. This matters here because during a controlled replay of the data and model handoff to a second engineering team, the following condition….
Incorrect Answers
A:Log reproducibility settings and rerun comparable trials with fixed split and controlled randomness before attributing changes. This is an action or diagnostic relevant to experiment random seed and data order variance, not one of the required two steps. In this scenario verify: Inspect judge rubric prompts, evaluation dataset balance, repeated ratings, human benchmark and uncertainty in pairwise preferences. Then correct: Build a constrained judge rubric with blind repeated comparisons and human-adjudicated calibration on representative examples.
C:Redact or tokenize unnecessary identifying fields before labeling and limit access to approved mappings. This is an action or diagnostic relevant to data minimization for labeling inputs, not one of the required two steps. In this scenario verify: Inspect judge rubric prompts, evaluation dataset balance, repeated ratings, human benchmark and uncertainty in pairwise preferences. Then correct: Build a constrained judge rubric with blind repeated comparisons and human-adjudicated calibration on representative examples.
D:Inspect Agent Platform ML Metadata artifacts, pipeline executions, dataset fingerprints and deployment revisions. This is an action or diagnostic relevant to ml metadata lineage for production model, not one of the required two steps. In this scenario verify: Inspect judge rubric prompts, evaluation dataset balance, repeated ratings, human benchmark and uncertainty in pairwise preferences. Then correct: Build a constrained judge rubric with blind repeated comparisons and human-adjudicated calibration on representative examples.
Question 786
While preparing a production training workflow on Gemini Enterprise Agent Platform, the following condition occurs: A product search team trains a binary spam classifier to find the most semantically relevant help article for a free-text question. Engineers need an auditable, resource-aware correction to the issue. What should the engineer inspect before selecting a correction?
- Inspect output schema, decoding configuration, structured-output support and malformed samples under representative prompts.
- Inspect task requirements for retrieval ranking, available query-document pairs, embedding similarity and labeled relevance examples.
- Examine model governance constraints, feature sensitivity, explanation fidelity and performance of simpler interpretable baselines.
- Inspect shutdown events, checkpoint directory, durable artifact transfers and restoration logic after simulated interruptions.
Explanation
Correct Answer
B:Inspect task requirements for retrieval ranking, available query-document pairs, embedding similarity and labeled relevance examples. This inspects evidence relevant to embedding retrieval versus text classifier. A binary classifier assigns predefined labels; semantic search needs query-to-document relevance ranking. After identifying the cause, the matching correction is: Use suitable text embeddings with vector similarity search and reranking rather than a binary spam decision. This matters here because while preparing a production training workflow on Gemini Enterprise Agent Platform, the following condition occurs:….
Incorrect Answers
A: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 embedding retrieval versus text classifier. The relevant evidence is: Inspect task requirements for retrieval ranking, available query-document pairs, embedding similarity and labeled relevance examples.
C: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 embedding retrieval versus text classifier. The relevant evidence is: Inspect task requirements for retrieval ranking, available query-document pairs, embedding similarity and labeled relevance examples. This matters here because while preparing a production training workflow on Gemini Enterprise Agent Platform, the following condition occurs:….
D:Inspect shutdown events, checkpoint directory, durable artifact transfers and restoration logic after simulated interruptions. This examines training checkpoint persistence across interruptions rather than the specific mechanism reported here. It is not the decisive first check for embedding retrieval versus text classifier. The relevant evidence is: Inspect task requirements for retrieval ranking, available query-document pairs, embedding similarity and labeled relevance examples.
Question 787
While reproducing a training orchestration incident in isolated staging, the following condition occurs: A Kubeflow preprocessing task is evicted repeatedly because it requests little memory despite materializing large embeddings inside the container. The team requires an idempotent pipeline behavior and traceable evidence. What should the engineer inspect before selecting a correction?
- Inspect the pipeline DAG, dependency declarations, artifact references and execution ordering from recent runs.
- Inspect Kubernetes task events, memory usage, component resource requests and container data-loading behavior.
- Compare serialized preprocessing artifacts, schema versions, train-time transformations and live request feature IDs.
- Inspect training data freshness, drift signals, label availability, business cadence and automated promotion rules.
Explanation
Correct Answer
B:Inspect Kubernetes task events, memory usage, component resource requests and container data-loading behavior. This inspects evidence relevant to custom kfp component resource limits. Kubernetes resource limits constrain component runtime even when the model's training job itself has spare capacity. After identifying the cause, the matching correction is: Tune component resource requests/limits and stream or partition embeddings rather than materializing every vector at once. This matters here because while reproducing a training orchestration incident in isolated staging, the following condition occurs: A Kubeflow….
Incorrect Answers
A:Inspect the pipeline DAG, dependency declarations, artifact references and execution ordering from recent runs. This examines pipeline component dependency order rather than the specific mechanism reported here. It is not the decisive first check for custom kfp component resource limits. The relevant evidence is: Inspect Kubernetes task events, memory usage, component resource requests and container data-loading behavior.
C:Compare serialized preprocessing artifacts, schema versions, train-time transformations and live request feature IDs. This examines training serving preprocessing artifact sharing rather than the specific mechanism reported here. It is not the decisive first check for custom kfp component resource limits. The relevant evidence is: Inspect Kubernetes task events, memory usage, component resource requests and container data-loading behavior.
D:Inspect training data freshness, drift signals, label availability, business cadence and automated promotion rules. This examines data drift triggered retraining policy rather than the specific mechanism reported here. It is not the decisive first check for custom kfp component resource limits. The relevant evidence is: Inspect Kubernetes task events, memory usage, component resource requests and container data-loading behavior.
Question 788
Within a jointly maintained Google Cloud data science workspace, the following condition occurs: An analyst runs a long-lived custom Apache Beam streaming service to compute a simple daily SQL join and aggregate over tables already in BigQuery. Collaborators must avoid expanding data access or breaking experiment reproducibility. Which configuration or procedure is the most appropriate correction?
- Implement the batch aggregate in scheduled BigQuery SQL when the required transformation is expressible and fits its SLA.
- Use a narrowly scoped service identity or approved user access, least-privilege dataset views and separated development data.
- Pin compatible dependency versions in a reproducible environment specification and validate fresh builds on clean runtimes.
- Standardize stable entity identifiers and validated feature definitions across training and online serving pathways.
Explanation
Correct Answer
A:Implement the batch aggregate in scheduled BigQuery SQL when the required transformation is expressible and fits its SLA. This is the targeted corrective response for bigquery versus dataflow preparation selection. Choosing a managed SQL engine for warehouse-resident batch joins avoids unnecessary always-on stream processing. Verify the diagnosis using: Examine source locations, batch cadence, SQL transform complexity, streaming latency requirements and operational burden. Confirm success using: The scheduled query produces equivalent daily features with less service orchestration and within the run window.
Incorrect Answers
B:Use a narrowly scoped service identity or approved user access, least-privilege dataset views and separated development data. This action addresses colab enterprise notebook data access scope, not the failure in the described bigquery versus dataflow preparation selection case. The targeted correction is: Implement the batch aggregate in scheduled BigQuery SQL when the required transformation is expressible and fits its SLA.
C:Pin compatible dependency versions in a reproducible environment specification and validate fresh builds on clean runtimes. This action addresses notebook dependency lock reproducibility, not the failure in the described bigquery versus dataflow preparation selection case. The targeted correction is: Implement the batch aggregate in scheduled BigQuery SQL when the required transformation is expressible and fits its SLA.
D:Standardize stable entity identifiers and validated feature definitions across training and online serving pathways. This action addresses feature store online and offline entity alignment, not the failure in the described bigquery versus dataflow preparation selection case. The targeted correction is: Implement the batch aggregate in scheduled BigQuery SQL when the required transformation is expressible and fits its SLA.
Question 789
During model-training validation on a Google Cloud experimental project, 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. The change should preserve a meaningful independent model-quality comparison. Which technical explanation best accounts for the reported behavior?
- Data parallel replicas each hold a model copy; they do not solve a model-too-large-for-one-device problem.
- Accuracy on highly imbalanced labels can hide severe minority-class recall failures.
- External reference updates need fresh context; changing foundation weights for every factual revision is inefficient.
- Structured numeric forecasting relies on temporal evidence; a conversational model is not automatically the best forecaster.
Explanation
Correct Answer
B:Accuracy on highly imbalanced labels can hide severe minority-class recall failures. This explains the mechanism behind cost-sensitive threshold for rare positives and why the described failure is possible. A discriminating investigation is: Inspect prevalence, confusion matrix, false-negative harm, calibration and precision-recall curves at candidate thresholds. The corresponding repair is: Choose a decision threshold using acceptable recall and alert burden with a validated independent cohort.
Incorrect Answers
A:Data parallel replicas each hold a model copy; they do not solve a model-too-large-for-one-device problem. This principle describes model parallelism for memory-bound transformer, which is not the cause of the observed cost-sensitive threshold for rare positives failure. The controlling explanation is: Accuracy on highly imbalanced labels can hide severe minority-class recall failures.
C:External reference updates need fresh context; changing foundation weights for every factual revision is inefficient. This principle describes foundation model prompt versus fine-tuning, which is not the cause of the observed cost-sensitive threshold for rare positives failure. The controlling explanation is: Accuracy on highly imbalanced labels can hide severe minority-class recall failures.
D:Structured numeric forecasting relies on temporal evidence; a conversational model is not automatically the best forecaster. This principle describes forecast trend and seasonality versus llm, which is not the cause of the observed cost-sensitive threshold for rare positives failure. The controlling explanation is: Accuracy on highly imbalanced labels can hide severe minority-class recall failures.
Question 790
During a postdeployment AI safety and reliability review, the following condition occurs: A service account running an AI workload can read all analytics datasets and copy confidential training data to an unrelated storage bucket. The model owner must provide repeatable evidence rather than a cosmetic dashboard change. Which TWO complementary choices jointly establish the evidence and appropriate correction? Select TWO.
- Examine explanation distributions, feature generation changes, protected attribute proxies and prediction outcomes by cohort.
- Apply supported safety filters and output policy checks with a controlled refusal or human review path for blocked results.
- Restrict dataset and bucket access by principal and resource; enforce approved egress controls and audit sensitive reads.
- Review IAM resource bindings, bucket access policies, egress paths and audit records for unauthorized data transfers.
- Version a single feature contract and add automated parity tests to the preprocessing pipeline and serving workflow.
Explanation
Correct Answers
C:Restrict dataset and bucket access by principal and resource; enforce approved egress controls and audit sensitive reads. This is the targeted corrective response for training data exfiltration through broad access. Least-privilege access and destination controls reduce unauthorized model-data disclosure risk. Verify the diagnosis using: Review IAM resource bindings, bucket access policies, egress paths and audit records for unauthorized data transfers. Confirm success using: The authorized pipeline still reads its intended data while attempts to export to unapproved destinations are denied.
D:Review IAM resource bindings, bucket access policies, egress paths and audit records for unauthorized data transfers. This inspects evidence relevant to training data exfiltration through broad access. Least-privilege access and destination controls reduce unauthorized model-data disclosure risk. After identifying the cause, the matching correction is: Restrict dataset and bucket access by principal and resource; enforce approved egress controls and audit sensitive reads.
Incorrect Answers
A:Examine explanation distributions, feature generation changes, protected attribute proxies and prediction outcomes by cohort. This is an action or diagnostic relevant to feature attribution drift after deployment, not one of the required two steps. In this scenario verify: Review IAM resource bindings, bucket access policies, egress paths and audit records for unauthorized data transfers. Then correct: Restrict dataset and bucket access by principal and resource; enforce approved egress controls and audit sensitive reads.
B:Apply supported safety filters and output policy checks with a controlled refusal or human review path for blocked results. This is an action or diagnostic relevant to gen ai output safety policy enforcement, not one of the required two steps. In this scenario verify: Review IAM resource bindings, bucket access policies, egress paths and audit records for unauthorized data transfers. Then correct: Restrict dataset and bucket access by principal and resource; enforce approved egress controls and audit sensitive reads.
E:Version a single feature contract and add automated parity tests to the preprocessing pipeline and serving workflow. This is an action or diagnostic relevant to training-serving feature skew detection, not one of the required two steps. In this scenario verify: Review IAM resource bindings, bucket access policies, egress paths and audit records for unauthorized data transfers. Then correct: Restrict dataset and bucket access by principal and resource; enforce approved egress controls and audit sensitive reads.