Google Professional Machine Learning Engineer Practice Test 13 - Prepaway
Practice Exams:

Pass Google Professional Machine Learning Engineer Exam in First Attempt Guaranteed!

Get 100% Latest Exam Questions, Accurate & Verified Answers!

30 Days Free Updates, Instant Download!

Verified By Experts
Premium Bundle
$29.98
$79.99 $109.97

Professional Machine Learning Engineer Premium Bundle


  • Premium File 402 Questions & Answers. Last update: Nov 13, 2026
  • Training Course 69 Lectures
  • Study Guide 376 Pages

Google Professional Machine Learning Engineer Practice Test 13

Question 120

At an energy distribution utility, a change advisory board is reviewing ML input and inference paths for metering and outage response applications. The reported issue is that a custom PyTorch training image works locally but fails in managed Agent Platform jobs after launch. Observation: cloud Logging shows arguments and input paths are ignored before the model writes checkpoints. The team must preserve approved data handling and avoid widening access to sensitive features. Which intervention most directly addresses the underlying Google Cloud ML workflow failure?

  1. Benchmark approved CPU GPU and TPU configurations at equivalent training targets including preprocessing costs
  2. Fit preprocessing only on training data and freeze transforms for all held-out cohorts
  3. Establish representative baseline evaluation and compare prompt grounding and lightweight adaptation before tuning
  4. Conform to the supported custom training container invocation and artifact path contract

Explanation

Correct Answer

D:

Conform to the supported custom training container invocation and artifact path contract. A custom PyTorch training image works locally but fails in managed Agent Platform jobs after launch. This matters because custom containers must satisfy the managed job runtime interface not only contain libraries.

Incorrect Answers

A:

Benchmark approved CPU GPU and TPU configurations at equivalent training targets including preprocessing costs. A custom PyTorch training image works locally but fails in managed Agent Platform jobs after launch. Conform to the supported custom training container invocation and artifact path contract. Custom containers must satisfy the managed job runtime interface not only contain libraries.

B:

Fit preprocessing only on training data and freeze transforms for all held-out cohorts. A custom PyTorch training image works locally but fails in managed Agent Platform jobs after launch. Conform to the supported custom training container invocation and artifact path contract. Custom containers must satisfy the managed job runtime interface not only contain libraries.

C:

Establish representative baseline evaluation and compare prompt grounding and lightweight adaptation before tuning. A custom PyTorch training image works locally but fails in managed Agent Platform jobs after launch. Conform to the supported custom training container invocation and artifact path contract.

Question 121

A controlled cloud model operations assessment for a business-to-business SaaS operator concerns tenant-specific document workflows. Reviewers find that a training feature pipeline launches Spark workers outside the permitted residency boundary after a default configuration change. The proposed remedy is to bind both processing location and storage to approved Regions with organizational policy checks. Any change must preserve the approved model contract and show measurable verification evidence. Which finding would best distinguish the suspected cause from nearby machine learning issues?

  1. Batch job location and Cloud Storage object audit reveal out-of-region reads for protected rows
  2. Event samples share stable transaction IDs and aggregate sums exceed the reconciled ledger
  3. Data classification scan identifies direct identifiers in text sent to embedding generation
  4. Query plan shows extensive scan despite efficient customer filtering within scanned blocks

Explanation

Correct Answer

A:

Batch job location and Cloud Storage object audit reveal out-of-region reads for protected rows. A training feature pipeline launches Spark workers outside the permitted residency boundary after a default configuration change. Regional placement is a data governance requirement across compute as well as storage. The correction is to bind both processing location and storage to approved Regions with organizational policy checks.

Incorrect Answers

B:

Event samples share stable transaction IDs and aggregate sums exceed the reconciled ledger. That does not establish the necessary result for a training feature pipeline launches Spark workers outside the permitted residency boundary after a default configuration change. In this case, regional placement is a data governance requirement across compute as well as storage. Use bind both processing location and storage to approved Regions with organizational policy checks.

C:

Data classification scan identifies direct identifiers in text sent to embedding generation. A training feature pipeline launches Spark workers outside the permitted residency boundary after a default configuration change. Batch job location and Cloud Storage object audit reveal out-of-region reads for protected rows.

D:

Query plan shows extensive scan despite efficient customer filtering within scanned blocks. A training feature pipeline launches Spark workers outside the permitted residency boundary after a default configuration change. Batch job location and Cloud Storage object audit reveal out-of-region reads for protected rows. Regional placement is a data governance requirement across compute as well as storage.

Question 122

At a healthcare claims processor, a change advisory board is reviewing ML input and inference paths for insurance enrollment and eligibility systems. The reported issue is that an internal Gemini assistant records full customer prompts including credentials and private notes in a broadly readable diagnostics sink. Confirming evidence shows that logging configuration and sample entries show secrets in plaintext outside approved retention scope. Any change must preserve the approved model contract and show measurable verification evidence. Which implementation paired with its verification outcome addresses the same fault?

  1. Minimize redact encrypt and restrict diagnostic data with an approved retention schedule — verify: Unsafe requests cause no side effects and permitted tasks have full auditability
  2. Minimize redact encrypt and restrict diagnostic data with an approved retention schedule — verify: No raw secrets appear in sampled logs and authorized investigation remains possible
  3. Minimize redact encrypt and restrict diagnostic data with an approved retention schedule — verify: All priority cohorts meet approved disparity and error constraints on independent holdouts
  4. Expand evaluation to multi-turn adversarial probes and enforce durable trusted policy context — verify: Known escalation dialogues are blocked without breaking legitimate conversations

Explanation

Correct Answer

B:

Minimize redact encrypt and restrict diagnostic data with an approved retention schedule. Verify that No raw secrets appear in sampled logs and authorized investigation remains possible. Logging configuration and sample entries show secrets in plaintext outside approved retention scope. Observability must not create a new sensitive-data exfiltration channel.

Incorrect Answers

A:

Minimize redact encrypt and restrict diagnostic data with an approved retention schedule. The check examines whether Unsafe requests cause no side effects and permitted tasks have full auditability. This check measures unsafe requests cause no side effects and permitted tasks have full auditability. Logging configuration and sample entries show secrets in plaintext outside approved retention scope. No raw secrets appear in sampled logs and authorized investigation remains possible.

C:

Minimize redact encrypt and restrict diagnostic data with an approved retention schedule. This check concerns All priority cohorts meet approved disparity and error constraints on independent holdouts, not the required outcome. Logging configuration and sample entries show secrets in plaintext outside approved retention scope. No raw secrets appear in sampled logs and authorized investigation remains possible.

D:

Expand evaluation to multi-turn adversarial probes and enforce durable trusted policy context. The check examines whether Known escalation dialogues are blocked without breaking legitimate conversations. Logging configuration and sample entries show secrets in plaintext outside approved retention scope. Minimize redact encrypt and restrict diagnostic data with an approved retention schedule. Verify that no raw secrets appear in sampled logs and authorized investigation remains possible.

Question 123

At an energy distribution utility, a change advisory board is reviewing ML input and inference paths for metering and outage response applications. The reported issue is that a training workflow fails on TPU after a code change introduces an operation without a compatible accelerator kernel. Confirming evidence shows that compiler diagnostic identifies an unsupported op while the same code succeeds on CPU. Only a controlled and reversible ML workflow change fits the approved release window. Which implementation paired with its verification outcome addresses the same fault?

  1. Choose a supported operator or compute target and test graph compilation before full runs — verify: Recomputed unbiased holdout metrics remain credible with no test information in transformations
  2. Choose a supported operator or compute target and test graph compilation before full runs — verify: Offline scores match features actually available during live decisions
  3. Choose a supported operator or compute target and test graph compilation before full runs — verify: A small accelerator smoke test passes and sustained throughput meets design requirements
  4. Shard and prefetch training records using an approved dataset layout and cache policy — verify: Input wait decreases and GPU utilization rises without missing image samples

Explanation

Correct Answer

C:

Choose a supported operator or compute target and test graph compilation before full runs. Verify that A small accelerator smoke test passes and sustained throughput meets design requirements. Compiler diagnostic identifies an unsupported op while the same code succeeds on CPU. This matters because hardware acceleration is constrained by framework operations and device compatibility.

Incorrect Answers

A:

Choose a supported operator or compute target and test graph compilation before full runs. This check concerns Recomputed unbiased holdout metrics remain credible with no test information in transformations, not the required outcome. Compiler diagnostic identifies an unsupported op while the same code succeeds on CPU. A small accelerator smoke test passes and sustained throughput meets design requirements.

B:

Choose a supported operator or compute target and test graph compilation before full runs. The check examines whether Offline scores match features actually available during live decisions. This check measures offline scores match features actually available during live decisions. Compiler diagnostic identifies an unsupported op while the same code succeeds on CPU. A small accelerator smoke test passes and sustained throughput meets design requirements.

D:

Shard and prefetch training records using an approved dataset layout and cache policy. The check examines whether Input wait decreases and GPU utilization rises without missing image samples. Compiler diagnostic identifies an unsupported op while the same code succeeds on CPU. Choose a supported operator or compute target and test graph compilation before full runs. Verify that a small accelerator smoke test passes and sustained throughput meets design requirements.

Question 124

At a shipping and logistics operator, a change advisory board is reviewing ML input and inference paths for shipment-demand models and warehouse metrics. The reported issue is that a research training pipeline begins while an upstream writer is still uploading image shards. Observation: input manifests reference missing object generations and partial payload checksums differ from expected. The recommendation must address the demonstrated ML failure without an unrelated pipeline redesign. Which intervention most directly addresses the underlying Google Cloud ML workflow failure?

  1. Declare a stable schema and normalize validated numeric fields in a controlled staging transformation
  2. Partition by event date where suitable and filter partitions explicitly before optional clustering
  3. Use a purpose-scoped service account and separate notebook runtime permissions by environment
  4. Publish immutable validated shards and only then release the approved versioned input manifest

Explanation

Correct Answer

D:

Publish immutable validated shards and only then release the approved versioned input manifest. A research training pipeline begins while an upstream writer is still uploading image shards. Dataset publication needs a verified atomic readiness signal. Confirm the repair by checking that every declared object is present with matching digest and full training reads pass.

Incorrect Answers

A:

Declare a stable schema and normalize validated numeric fields in a controlled staging transformation. A research training pipeline begins while an upstream writer is still uploading image shards. In this case, dataset publication needs a verified atomic readiness signal. Publish immutable validated shards and only then release the approved versioned input manifest.

B:

Partition by event date where suitable and filter partitions explicitly before optional clustering. A research training pipeline begins while an upstream writer is still uploading image shards. Publish immutable validated shards and only then release the approved versioned input manifest.

C:

Use a purpose-scoped service account and separate notebook runtime permissions by environment. A research training pipeline begins while an upstream writer is still uploading image shards. Publish immutable validated shards and only then release the approved versioned input manifest. Dataset publication needs a verified atomic readiness signal.

Question 125

A controlled cloud model operations assessment for a global online retailer concerns purchase forecasting and recommendation endpoints. Reviewers find that a pipeline reuses a cached transform despite a changed schema contract because cache identity excludes config version. Confirming evidence shows that cached artifact metadata shows old feature schema while current input version changed. Service owners require a reproducible result without unsafe or unapproved production testing. Which implementation paired with its verification outcome addresses the same fault?

  1. Include relevant schema and transformation configuration in cache keys or invalidate results — verify: Changed configs force correct processing and identical configs reproduce cached output
  2. Include relevant schema and transformation configuration in cache keys or invalidate results — verify: Cross-tenant read attempts fail while authorized pipelines retain artifact lineage
  3. Include relevant schema and transformation configuration in cache keys or invalidate results — verify: Job retries yield one approved release with consistent metadata and no orphan deployments
  4. Use point-in-time feature availability and isolated future-time holdouts in training automation — verify: Promotion uses unbiased metrics from data truly available at inference

Explanation

Correct Answer

A:

Include relevant schema and transformation configuration in cache keys or invalidate results. Verify that Changed configs force correct processing and identical configs reproduce cached output. Cached artifact metadata shows old feature schema while current input version changed. Pipeline caching is safe only when cache identity includes all behavior-changing inputs.

Incorrect Answers

B:

Include relevant schema and transformation configuration in cache keys or invalidate results. This check concerns Cross-tenant read attempts fail while authorized pipelines retain artifact lineage, not the required outcome. Cached artifact metadata shows old feature schema while current input version changed. Changed configs force correct processing and identical configs reproduce cached output.

C:

Include relevant schema and transformation configuration in cache keys or invalidate results. The check examines whether Job retries yield one approved release with consistent metadata and no orphan deployments. This check measures job retries yield one approved release with consistent metadata and no orphan deployments. Cached artifact metadata shows old feature schema while current input version changed. Changed configs force correct processing and identical configs reproduce cached output.

D:

Use point-in-time feature availability and isolated future-time holdouts in training automation. The check examines whether Promotion uses unbiased metrics from data truly available at inference. Cached artifact metadata shows old feature schema while current input version changed. Include relevant schema and transformation configuration in cache keys or invalidate results. Verify that changed configs force correct processing and identical configs reproduce cached output.

Question 126

At an energy distribution utility, a change advisory board is reviewing ML input and inference paths for metering and outage response applications. The reported issue is that an approved model version gets a newly deployed preprocessing transform without review, altering outputs after a no-op rollout. Confirming evidence shows that artifact provenance shows feature code version changed while model weights remain identical. Any change must preserve the approved model contract and show measurable verification evidence. Which implementation paired with its verification outcome addresses the same fault?

  1. Promote model and feature transforms as a single immutable tested deployment unit — verify: Representative requests complete with bounded retries and no partial duplicate updates
  2. Promote model and feature transforms as a single immutable tested deployment unit — verify: The identical release yields identical transformed features and model predictions on replay
  3. Promote model and feature transforms as a single immutable tested deployment unit — verify: Smoke tests pass representative request formats and prediction values against reference
  4. Separate serving and training service accounts and enforce least privilege on data stores — verify: Serving can read only approved model and online features while historic labels are denied

Explanation

Correct Answer

B:

Promote model and feature transforms as a single immutable tested deployment unit. Verify that The identical release yields identical transformed features and model predictions on replay. Artifact provenance shows feature code version changed while model weights remain identical. This matters because an ML serving release encompasses both transformations and learned parameters.

Incorrect Answers

A:

Promote model and feature transforms as a single immutable tested deployment unit. The check examines whether Representative requests complete with bounded retries and no partial duplicate updates. This check measures representative requests complete with bounded retries and no partial duplicate updates. Artifact provenance shows feature code version changed while model weights remain identical. The identical release yields identical transformed features and model predictions on replay.

C:

Promote model and feature transforms as a single immutable tested deployment unit. The check examines whether Smoke tests pass representative request formats and prediction values against reference. This check measures smoke tests pass representative request formats and prediction values against reference. Artifact provenance shows feature code version changed while model weights remain identical. The identical release yields identical transformed features and model predictions on replay.

D:

Separate serving and training service accounts and enforce least privilege on data stores. The check examines whether Serving can read only approved model and online features while historic labels are denied. Artifact provenance shows feature code version changed while model weights remain identical. Promote model and feature transforms as a single immutable tested deployment unit. Verify that the identical release yields identical transformed features and model predictions on replay.

Question 127

At a global online retailer, a change advisory board is reviewing ML input and inference paths for purchase forecasting and recommendation endpoints. The reported issue is that a classification pipeline replaces missing income with zero and wrongly treats undocumented values as verified low income. The team proposes to define separate missingness treatment and an imputation policy fitted on training data. Only a controlled and reversible ML workflow change fits the approved release window. Which post-change observation would establish the required ML or data quality outcome?

  1. Training succeeds using approved features and denied columns cannot be read by the job principal
  2. Privacy tests on the versioned tuning dataset pass without degrading approved task coverage
  3. Unknown and genuine zero values remain distinguishable and holdout calibration improves
  4. No account overlaps final splits and deployment-shaped scores remain within thresholds

Explanation

Correct Answer

C:

Unknown and genuine zero values remain distinguishable and holdout calibration improves. A classification pipeline replaces missing income with zero and wrongly treats undocumented values as verified low income. Missing data must preserve semantics instead of inventing measured values. The correction is to define separate missingness treatment and an imputation policy fitted on training data.

Incorrect Answers

A:

Training succeeds using approved features and denied columns cannot be read by the job principal. A classification pipeline replaces missing income with zero and wrongly treats undocumented values as verified low income. Unknown and genuine zero values remain distinguishable and holdout calibration improves.

B:

Privacy tests on the versioned tuning dataset pass without degrading approved task coverage. A classification pipeline replaces missing income with zero and wrongly treats undocumented values as verified low income. Unknown and genuine zero values remain distinguishable and holdout calibration improves. Missing data must preserve semantics instead of inventing measured values.

D:

No account overlaps final splits and deployment-shaped scores remain within thresholds. A classification pipeline replaces missing income with zero and wrongly treats undocumented values as verified low income. Unknown and genuine zero values remain distinguishable and holdout calibration improves.

Question 128

At a financial data provider, a change advisory board is reviewing ML input and inference paths for internal analytics and partner market feeds. The reported issue is that a deployment task runs before model-quality approval because the DAG does not express the prerequisite edge. The proposed remedy is to enforce explicit validation and human approval dependencies in the deployment graph. The recommendation must address the demonstrated ML failure without an unrelated pipeline redesign. Which finding would best distinguish the suspected cause from nearby machine learning issues?

  1. Model metadata shows baseline URI points to a staging registry version
  2. Job history shows train start precedes finalized partition events and record counts vary by retry
  3. Feature lineage shows a status timestamp later than the intended prediction time
  4. Task execution graph shows deployment and validation run in parallel with production mutations

Explanation

Correct Answer

D:

Task execution graph shows deployment and validation run in parallel with production mutations. A deployment task runs before model-quality approval because the DAG does not express the prerequisite edge. Workflow task ordering must encode governance dependencies not rely on scheduling coincidence. The correction is to enforce explicit validation and human approval dependencies in the deployment graph.

Incorrect Answers

A:

Model metadata shows baseline URI points to a staging registry version. A deployment task runs before model-quality approval because the DAG does not express the prerequisite edge. Task execution graph shows deployment and validation run in parallel with production mutations. Workflow task ordering must encode governance dependencies not rely on scheduling coincidence.

B:

Job history shows train start precedes finalized partition events and record counts vary by retry. A deployment task runs before model-quality approval because the DAG does not express the prerequisite edge. Task execution graph shows deployment and validation run in parallel with production mutations.

C:

Feature lineage shows a status timestamp later than the intended prediction time. That does not establish the necessary result for a deployment task runs before model-quality approval because the DAG does not express the prerequisite edge. In this case, workflow task ordering must encode governance dependencies not rely on scheduling coincidence. Use enforce explicit validation and human approval dependencies in the deployment graph.

Question 129

At a manufacturing supply-chain platform, a change advisory board is reviewing ML input and inference paths for supplier ordering and inventory reconciliation. The reported issue is that a transient Cloud Build failure retries the publish step and writes two different registry versions for one approved dataset. The relevant evidence is that registry history shows repeated version creation with identical source manifest but different deployment ordering. The plan is to use deterministic run IDs and idempotent model-registration checks before promotion. Which machine learning principle best explains the chosen response?

  1. Retry-safe CI/CD requires idempotent artifact promotion semantics
  2. Continuous evaluation should balance current deployment populations with regression stability
  3. Safe ML delivery needs a separately protected rollback artifact identity
  4. CI/CD pipelines must verify the exact software and model payload they promote

Explanation

Correct Answer

A:

Retry-safe CI/CD requires idempotent artifact promotion semantics. Registry history shows repeated version creation with identical source manifest but different deployment ordering. The correction is to use deterministic run IDs and idempotent model-registration checks before promotion.

Incorrect Answers

B:

Continuous evaluation should balance current deployment populations with regression stability. Registry history shows repeated version creation with identical source manifest but different deployment ordering. Retry-safe CI/CD requires idempotent artifact promotion semantics.

C:

Safe ML delivery needs a separately protected rollback artifact identity. Registry history shows repeated version creation with identical source manifest but different deployment ordering. Retry-safe CI/CD requires idempotent artifact promotion semantics.

D:

CI/CD pipelines must verify the exact software and model payload they promote. Registry history shows repeated version creation with identical source manifest but different deployment ordering. Retry-safe CI/CD requires idempotent artifact promotion semantics.

Question 130

A controlled cloud model operations assessment for a food distribution business concerns cold-chain reporting and store replenishment. Reviewers find that a warehouse alert system trains a costly neural network to flag static rule violations already defined exactly in policy. Observation: requirement analysis shows outputs are deterministic Boolean constraints and no uncertain pattern needs learning. Any change must preserve the approved model contract and show measurable verification evidence. Which intervention most directly addresses the underlying Google Cloud ML workflow failure?

  1. Evaluate suitable ranking models using top-k relevance metrics and guardrails
  2. Implement governed SQL or rules where suitable instead of unnecessary custom training
  3. Compare time-aware forecast approaches with legitimate calendar features and rolling backtests
  4. Choose a supported synchronized distributed training strategy and validate gradient aggregation

Explanation

Correct Answer

B:

Implement governed SQL or rules where suitable instead of unnecessary custom training. A warehouse alert system trains a costly neural network to flag static rule violations already defined exactly in policy. Not every data decision warrants ML when exact deterministic rules solve the requirement. Confirm the repair by checking that rule checks reproduce expected decisions with lower latency and auditable business logic.

Incorrect Answers

A:

Evaluate suitable ranking models using top-k relevance metrics and guardrails. A warehouse alert system trains a costly neural network to flag static rule violations already defined exactly in policy. Implement governed SQL or rules where suitable instead of unnecessary custom training. Not every data decision warrants ML when exact deterministic rules solve the requirement.

C:

Compare time-aware forecast approaches with legitimate calendar features and rolling backtests. A warehouse alert system trains a costly neural network to flag static rule violations already defined exactly in policy. Implement governed SQL or rules where suitable instead of unnecessary custom training.

D:

Choose a supported synchronized distributed training strategy and validate gradient aggregation. A warehouse alert system trains a costly neural network to flag static rule violations already defined exactly in policy. In this case, not every data decision warrants ML when exact deterministic rules solve the requirement. Implement governed SQL or rules where suitable instead of unnecessary custom training.

Professional Machine Learning Engineer Practice Tests