Microsoft AI-300 Practice Test 72
Question 710
During a production Azure Machine Learning model lifecycle incident, the following condition occurs: A nightly 15-million-row customer scoring task is sent as individual requests to a production real-time endpoint, degrading its daytime customer traffic. The operator must safeguard training reproducibility and serving continuity. Which production-safe remediation should the engineer apply to this AI workflow?
- Move the approved bulk workload to a managed batch endpoint or batch inference job while preserving online capacity for interactive calls.
- Restore the complete previously approved deployment package, including model, environment and scoring code, rather than only changing weights.
- Restrict access, redact or aggregate sensitive examples and publish a privacy-reviewed evaluation artifact for authorized reviewers.
- Align the published endpoint contract and client payload mapping, then deploy the reviewed scoring code and test representative requests.
Explanation
Correct Answer
A:Move the approved bulk workload to a managed batch endpoint or batch inference job while preserving online capacity for interactive calls. This targets overnight bulk scoring saturates interactive online endpoint. Batch scoring and synchronous online inference serve different workload shapes and should not contend for capacity unnecessarily. Verify the failure through: Inspect request volume, concurrency, endpoint latency graphs, job timing and batch-friendly data path availability. Validate success with: Overnight processing completes within its processing window and interactive endpoint latency remains within SLA. This matters here because during a production Azure Machine Learning model lifecycle incident, the following condition occurs: A nightly….
Incorrect Answers
B:Restore the complete previously approved deployment package, including model, environment and scoring code, rather than only changing weights. This could address rollback cannot recover a previous environment version, but it does not solve the specified overnight bulk scoring saturates interactive online endpoint failure. The matching remediation is: Move the approved bulk workload to a managed batch endpoint or batch inference job while preserving online capacity for interactive calls.
C:Restrict access, redact or aggregate sensitive examples and publish a privacy-reviewed evaluation artifact for authorized reviewers. This could address model evaluation exposes sensitive examples in explainability report, but it does not solve the specified overnight bulk scoring saturates interactive online endpoint failure. The matching remediation is: Move the approved bulk workload to a managed batch endpoint or batch inference job while preserving online capacity for interactive calls.
D:Align the published endpoint contract and client payload mapping, then deploy the reviewed scoring code and test representative requests. This could address endpoint scoring script expects wrong json schema, but it does not solve the specified overnight bulk scoring saturates interactive online endpoint failure. The matching remediation is: Move the approved bulk workload to a managed batch endpoint or batch inference job while preserving online capacity for interactive calls.
Question 711
During a regulated Azure ML workspace infrastructure review, the following condition occurs: A training job reads a similarly named blob container containing last quarter's features because its Azure ML datastore URI targets the staging location rather than the approved production store. The team must preserve approved access, model provenance and private network constraints. What should the engineer examine first to confirm the specific AI-300 workload issue?
- Inspect compute cluster requested SKU, max nodes, available family quota, current utilization, node provisioning events and region availability.
- Inspect az account subscription context, deployment scope flags, tenant identity, Bicep parameter file and cloud audit evidence.
- Inspect the workload principal, Azure ML workspace role assignment, subscription inheritance, required model-write actions and audit activity.
- Compare the datastore URI and backing storage container, recorded data asset path, job input mount path and expected data version checksum.
Explanation
Correct Answer
D:Compare the datastore URI and backing storage container, recorded data asset path, job input mount path and expected data version checksum. This is the most relevant evidence for datastore uri selects the incorrect storage container. A datastore connection and data asset version identify where training inputs come from; paths must be validated independently from authentication success. The matching controlled remediation is: Update the datastore or data asset binding to the governed storage path with an explicitly pinned version and validated read access. This matters here because a training job reads a similarly named blob container containing last quarter's features because its Azure ML….
Incorrect Answers
A:Inspect compute cluster requested SKU, max nodes, available family quota, current utilization, node provisioning events and region availability. This would investigate compute cluster cannot allocate requested gpu quota rather than the documented datastore uri selects the incorrect storage container issue. The relevant investigation is: Compare the datastore URI and backing storage container, recorded data asset path, job input mount path and expected data version checksum.
B:Inspect az account subscription context, deployment scope flags, tenant identity, Bicep parameter file and cloud audit evidence. This would investigate azure cli deployment silently targets sandbox defaults rather than the documented datastore uri selects the incorrect storage container issue. The relevant investigation is: Compare the datastore URI and backing storage container, recorded data asset path, job input mount path and expected data version checksum.
C:Inspect the workload principal, Azure ML workspace role assignment, subscription inheritance, required model-write actions and audit activity. This would investigate workspace contributor role granted to untrusted automation rather than the documented datastore uri selects the incorrect storage container issue. The relevant investigation is: Compare the datastore URI and backing storage container, recorded data asset path, job input mount path and expected data version checksum.
Question 712
While deploying reproducible Azure Machine Learning infrastructure, the following condition occurs: A Bicep deployment provisions an Azure ML workspace but its associated storage, Key Vault and private endpoints are created after the workspace references them, producing intermittent deployment failures. The change must remain limited to authorized workspaces and avoid unreviewed production impact. What should the engineer examine first to confirm the specific AI-300 workload issue?
- Compare compute-instance schedules, last interactive activity, active notebook kernels, managed identity dependencies and cost reports.
- Examine workspace managed network isolation mode, outbound rules, registry destination, DNS and job network diagnostics.
- Inspect az account subscription context, deployment scope flags, tenant identity, Bicep parameter file and cloud audit evidence.
- Inspect Azure deployment operations, Bicep resource references, implicit dependency graph, private endpoint target IDs and pipeline parameter order.
Explanation
Correct Answer
D:Inspect Azure deployment operations, Bicep resource references, implicit dependency graph, private endpoint target IDs and pipeline parameter order. This is the most relevant evidence for bicep workspace declaration misses dependent private resources. Declarative infrastructure still requires a valid resource dependency graph and stable networking prerequisites; syntax validation alone does not enforce desired provisioning order. The matching controlled remediation is: Declare proper resource references or explicit dependencies and staged network prerequisites in the reviewed Bicep deployment. This matters here because while deploying reproducible Azure Machine Learning infrastructure, the following condition occurs: A Bicep….
Incorrect Answers
A:Compare compute-instance schedules, last interactive activity, active notebook kernels, managed identity dependencies and cost reports. This would investigate interactive compute instances remain running while unused rather than the documented bicep workspace declaration misses dependent private resources issue. The relevant investigation is: Inspect Azure deployment operations, Bicep resource references, implicit dependency graph, private endpoint target IDs and pipeline parameter order.
B:Examine workspace managed network isolation mode, outbound rules, registry destination, DNS and job network diagnostics. This would investigate aml managed network blocks model registry egress rather than the documented bicep workspace declaration misses dependent private resources issue. The relevant investigation is: Inspect Azure deployment operations, Bicep resource references, implicit dependency graph, private endpoint target IDs and pipeline parameter order.
C:Inspect az account subscription context, deployment scope flags, tenant identity, Bicep parameter file and cloud audit evidence. This would investigate azure cli deployment silently targets sandbox defaults rather than the documented bicep workspace declaration misses dependent private resources issue. The relevant investigation is: Inspect Azure deployment operations, Bicep resource references, implicit dependency graph, private endpoint target IDs and pipeline parameter order.
Question 713
During a Foundry generative AI quality and observability incident, the following condition occurs: A team upgrades a custom evaluator prompt and compares this month's agent scores to last month's, even though the grading rubric and baseline model changed. The fix must preserve evaluation validity and avoid unnecessary user-data collection. Which targeted Azure AI operations change best addresses the documented problem?
- Correct the test dataset mapping for query, context, response and reference fields and validate a sampled evaluation before the full run.
- Version the custom evaluator and recalibrate on a fixed reference set before using its scores for cross-version comparison.
- Use relevant retrieval context, justified response limits and supported caching after verifying quality on evaluation examples.
- Add approved prompt-injection and tool-abuse evaluation cases with tool permission gates and block release on unsafe execution.
Explanation
Correct Answer
B:Version the custom evaluator and recalibrate on a fixed reference set before using its scores for cross-version comparison. This targets custom evaluator silently changes its scoring rubric. Evaluation metrics are outputs of versioned assessment logic; changing a judge or rubric can change scores without any underlying agent change. Verify the failure through: Review evaluator code and prompt versions, rubric labels, baseline sample scores and published metric time series. Validate success with: Historical baselines can be reproduced with the same evaluator version and new rubric results are labeled separately.
Incorrect Answers
A:Correct the test dataset mapping for query, context, response and reference fields and validate a sampled evaluation before the full run. This could address genai evaluation dataset lacks expected answer column mapping, but it does not solve the specified custom evaluator silently changes its scoring rubric failure. The matching remediation is: Version the custom evaluator and recalibrate on a fixed reference set before using its scores for cross-version comparison.
C:Use relevant retrieval context, justified response limits and supported caching after verifying quality on evaluation examples. This could address prompt truncation strategy increases generated-token spend, but it does not solve the specified custom evaluator silently changes its scoring rubric failure. The matching remediation is: Version the custom evaluator and recalibrate on a fixed reference set before using its scores for cross-version comparison.
D:Add approved prompt-injection and tool-abuse evaluation cases with tool permission gates and block release on unsafe execution. This could address content safety evaluation omits prompt injection tests, but it does not solve the specified custom evaluator silently changes its scoring rubric failure. The matching remediation is: Version the custom evaluator and recalibrate on a fixed reference set before using its scores for cross-version comparison.
Question 714
During a production Azure Machine Learning model lifecycle incident, the following condition occurs: A scheduled Azure ML pipeline appears fast but returns last week's encoded features because a preprocessing component is reused under inputs that do not capture the changed transformation configuration. The operator must safeguard training reproducibility and serving continuity. Which TWO complementary choices identify the issue and correct it? Select TWO.
- Version the transformation component and inputs explicitly and disable or invalidate reuse where the effective computation has changed.
- Monitor available proxy signals immediately and compute labeled performance only after verified outcomes arrive through a governed join.
- Configure a relevant supported primary metric and validation scheme for rare positive events and compare operating thresholds after training.
- Inspect component reuse settings, input asset version, code and environment fingerprints, effective parameters and cached job lineage.
- Review distributed sampler configuration, global and local ranks, dataset partition checks, epoch seeding and examples processed per worker.
Explanation
Correct Answers
A:Version the transformation component and inputs explicitly and disable or invalidate reuse where the effective computation has changed. This targets pipeline cache reuses stale transform artifact. Pipeline reuse is correct only when the declared inputs and environment represent all behaviorally relevant computation dependencies. Verify the failure through: Inspect component reuse settings, input asset version, code and environment fingerprints, effective parameters and cached job lineage. Validate success with: A rerun executes the revised transform and its output lineage identifies the current code and dataset rather than a stale cached job.
D:Inspect component reuse settings, input asset version, code and environment fingerprints, effective parameters and cached job lineage. This is the most relevant evidence for pipeline cache reuses stale transform artifact. Pipeline reuse is correct only when the declared inputs and environment represent all behaviorally relevant computation dependencies. The matching controlled remediation is: Version the transformation component and inputs explicitly and disable or invalidate reuse where the effective computation has changed.
Incorrect Answers
B:Monitor available proxy signals immediately and compute labeled performance only after verified outcomes arrive through a governed join. This is associated with accuracy monitoring lacks delayed outcome labels, not the essential evidence-and-fix pair for pipeline cache reuses stale transform artifact. First examine: Inspect component reuse settings, input asset version, code and environment fingerprints, effective parameters and cached job lineage. Then correct: Version the transformation component and inputs explicitly and disable or invalidate reuse where the effective computation has changed.
C:Configure a relevant supported primary metric and validation scheme for rare positive events and compare operating thresholds after training. This is associated with automl primary metric does not match rare-event goal, not the essential evidence-and-fix pair for pipeline cache reuses stale transform artifact. First examine: Inspect component reuse settings, input asset version, code and environment fingerprints, effective parameters and cached job lineage. Then correct: Version the transformation component and inputs explicitly and disable or invalidate reuse where the effective computation has changed.
E:Review distributed sampler configuration, global and local ranks, dataset partition checks, epoch seeding and examples processed per worker. This is associated with distributed training uses one shard for every worker, not the essential evidence-and-fix pair for pipeline cache reuses stale transform artifact. First examine: Inspect component reuse settings, input asset version, code and environment fingerprints, effective parameters and cached job lineage. Then correct: Version the transformation component and inputs explicitly and disable or invalidate reuse where the effective computation has changed.
Question 715
As an MLOps engineer reviews a model training or inference pipeline, the following condition occurs: An Azure ML notebook cell executes successfully after an ad hoc pip install, but the same code submitted as a managed job fails because the registered environment excludes that package. The resolution must preserve quality gates, documented artifacts and rollback options. Which investigation best verifies the cause before changing production configuration?
- Compare notebook kernel package list, managed job environment asset, dependency lock file and the remote import traceback.
- Review training and validation learning curves, selection rule, dataset splits and registered model candidate metrics.
- Inspect MLflow tracking URI and experiment name, run parameter history, metric logs, artifact URIs and the input data version of each job.
- Inspect request volume, concurrency, endpoint latency graphs, job timing and batch-friendly data path availability.
Explanation
Correct Answer
A:Compare notebook kernel package list, managed job environment asset, dependency lock file and the remote import traceback. This is the most relevant evidence for notebook experiment installs packages only in interactive kernel. Interactive notebook state can diverge from a managed training environment; reproducibility needs explicit environment assets. The matching controlled remediation is: Add the vetted package to a versioned Azure ML environment and submit the notebook's training logic as a reproducible job. This matters here because as an MLOps engineer reviews a model training or inference pipeline, the following condition occurs: An Azure ML….
Incorrect Answers
B:Review training and validation learning curves, selection rule, dataset splits and registered model candidate metrics. This would investigate new candidate chosen from training instead of validation results rather than the documented notebook experiment installs packages only in interactive kernel issue. The relevant investigation is: Compare notebook kernel package list, managed job environment asset, dependency lock file and the remote import traceback. This matters here because as an MLOps engineer reviews a model training or inference pipeline, the following condition occurs: An Azure ML….
C:Inspect MLflow tracking URI and experiment name, run parameter history, metric logs, artifact URIs and the input data version of each job. This would investigate mlflow experiment logs omit run parameters rather than the documented notebook experiment installs packages only in interactive kernel issue. The relevant investigation is: Compare notebook kernel package list, managed job environment asset, dependency lock file and the remote import traceback.
D:Inspect request volume, concurrency, endpoint latency graphs, job timing and batch-friendly data path availability. This would investigate overnight bulk scoring saturates interactive online endpoint rather than the documented notebook experiment installs packages only in interactive kernel issue. The relevant investigation is: Compare notebook kernel package list, managed job environment asset, dependency lock file and the remote import traceback.
Question 716
During a Foundry generative AI quality and observability incident, the following condition occurs: A production agent serves thousands of requests but its evaluation monitor collects only successful English FAQ interactions because failed sessions and other languages were excluded. The fix must preserve evaluation validity and avoid unnecessary user-data collection. Which production-safe remediation should the engineer apply to this AI workflow?
- Apply privacy-conscious representative sampling or stratified monitoring that includes error and underrepresented user cohorts.
- Segment continuous evaluation by model/agent deployment version and configure a focused regression alert for the candidate.
- Include appropriate complete conversation histories and tool events in the supported agent evaluation workflow while preserving privacy protections.
- Use relevant retrieval context, justified response limits and supported caching after verifying quality on evaluation examples.
Explanation
Correct Answer
A:Apply privacy-conscious representative sampling or stratified monitoring that includes error and underrepresented user cohorts. This targets continuous monitoring has no representative sampling coverage. Production GenAI quality monitoring is biased when sampling systematically excludes difficult or unsuccessful interactions. Verify the failure through: Inspect monitoring sample filters, event statuses, language distribution, evaluator eligibility and collected trace metadata. Validate success with: The monitoring dataset reflects significant live workflows and reveals failures in the previously omitted segments. This matters here because a production agent serves thousands of requests but its evaluation monitor collects only successful English FAQ….
Incorrect Answers
B:Segment continuous evaluation by model/agent deployment version and configure a focused regression alert for the candidate. This could address foundry continuous evaluation ignores production deployment version, but it does not solve the specified continuous monitoring has no representative sampling coverage failure. The matching remediation is: Apply privacy-conscious representative sampling or stratified monitoring that includes error and underrepresented user cohorts. This matters here because a production agent serves thousands of requests but its evaluation monitor collects only successful English FAQ….
C:Include appropriate complete conversation histories and tool events in the supported agent evaluation workflow while preserving privacy protections. This could address multi-turn agent evaluation uses isolated last message only, but it does not solve the specified continuous monitoring has no representative sampling coverage failure. The matching remediation is: Apply privacy-conscious representative sampling or stratified monitoring that includes error and underrepresented user cohorts.
D:Use relevant retrieval context, justified response limits and supported caching after verifying quality on evaluation examples. This could address prompt truncation strategy increases generated-token spend, but it does not solve the specified continuous monitoring has no representative sampling coverage failure. The matching remediation is: Apply privacy-conscious representative sampling or stratified monitoring that includes error and underrepresented user cohorts.
Question 717
During a Microsoft Foundry generative AI deployment review, the following condition occurs: A GenAI team commits real customer tickets containing identifiers as few-shot prompt examples in a shared Git repository, exposing data beyond the approved processing boundary. The deployment must retain governed identities, endpoint contracts and auditable release history. What approved correction resolves the failure without modifying unrelated healthy services?
- Store reviewed prompt templates and configuration in version control and deploy immutable approved revisions through the release workflow.
- Define the model and project deployment lifecycle dependencies and verify provisioning completion before dependent app configuration.
- Remove direct identifiers from ongoing prompt examples, rotate affected credentials if necessary and use sanitized synthetic cases under review.
- Define a versioned structured response instruction and supported JSON schema contract with parser-based validation.
Explanation
Correct Answer
C:Remove direct identifiers from ongoing prompt examples, rotate affected credentials if necessary and use sanitized synthetic cases under review. This targets prompt repository references secret customer examples. Source control is not an appropriate unreviewed store for sensitive customer data; prompt versioning requires privacy-preserving examples and access governance. Verify the failure through: Inspect prompt repository history, example dataset provenance, PII classification, branch access controls and affected downstream forks. Validate success with: Prompt tests retain representative behavior without exposing customer identities in accessible Git history or future deployments. This matters here because a GenAI team commits real customer tickets containing identifiers as few-shot prompt examples in a shared Git….
Incorrect Answers
A:Store reviewed prompt templates and configuration in version control and deploy immutable approved revisions through the release workflow. This could address prompt hotfix bypasses git review and loses rollback history, but it does not solve the specified prompt repository references secret customer examples failure. The matching remediation is: Remove direct identifiers from ongoing prompt examples, rotate affected credentials if necessary and use sanitized synthetic cases under review.
B:Define the model and project deployment lifecycle dependencies and verify provisioning completion before dependent app configuration. This could address foundry bicep rollout omits model deployment dependencies, but it does not solve the specified prompt repository references secret customer examples failure. The matching remediation is: Remove direct identifiers from ongoing prompt examples, rotate affected credentials if necessary and use sanitized synthetic cases under review.
D:Define a versioned structured response instruction and supported JSON schema contract with parser-based validation. This could address prompt omits explicit structured answer schema, but it does not solve the specified prompt repository references secret customer examples failure. The matching remediation is: Remove direct identifiers from ongoing prompt examples, rotate affected credentials if necessary and use sanitized synthetic cases under review.
Question 718
As an AI operations team checks production agent evaluation evidence, the following condition occurs: A GenAIOps release pipeline writes evaluation scores to a report but deploys an agent even when groundedness falls below the required threshold. The workflow must produce measurable quality, safety and performance evidence. Select TWO: the strongest initial evidence and the aligned controlled correction.
- Extend safety testing to representative supported languages and calibrate policy thresholds without weakening necessary allowed content.
- Examine evaluation records, chat history serialization, tool traces, single-turn versus multi-turn settings and representative failure conversations.
- Make required quality/safety criteria explicit CI gates and stop deployment when evaluated scores fail the approved tolerances.
- Use relevant retrieval context, justified response limits and supported caching after verifying quality on evaluation examples.
- Inspect evaluation workflow outputs, release conditions, metric thresholds, CI job failure behavior and promotion records.
Explanation
Correct Answers
C:Make required quality/safety criteria explicit CI gates and stop deployment when evaluated scores fail the approved tolerances. This targets evaluation workflow fails to block degraded agent deployment. Automated evaluation provides protection only when results are connected to enforceable deployment decisions. Verify the failure through: Inspect evaluation workflow outputs, release conditions, metric thresholds, CI job failure behavior and promotion records. Validate success with: An intentionally degraded candidate fails before production promotion and a passing candidate proceeds with an auditable evaluation receipt.
E:Inspect evaluation workflow outputs, release conditions, metric thresholds, CI job failure behavior and promotion records. This is the most relevant evidence for evaluation workflow fails to block degraded agent deployment. Automated evaluation provides protection only when results are connected to enforceable deployment decisions. The matching controlled remediation is: Make required quality/safety criteria explicit CI gates and stop deployment when evaluated scores fail the approved tolerances.
Incorrect Answers
A:Extend safety testing to representative supported languages and calibrate policy thresholds without weakening necessary allowed content. This is associated with harmful content probe set excludes multilingual input, not the essential evidence-and-fix pair for evaluation workflow fails to block degraded agent deployment. First examine: Inspect evaluation workflow outputs, release conditions, metric thresholds, CI job failure behavior and promotion records. Then correct: Make required quality/safety criteria explicit CI gates and stop deployment when evaluated scores fail the approved tolerances.
B:Examine evaluation records, chat history serialization, tool traces, single-turn versus multi-turn settings and representative failure conversations. This is associated with multi-turn agent evaluation uses isolated last message only, not the essential evidence-and-fix pair for evaluation workflow fails to block degraded agent deployment. First examine: Inspect evaluation workflow outputs, release conditions, metric thresholds, CI job failure behavior and promotion records. Then correct: Make required quality/safety criteria explicit CI gates and stop deployment when evaluated scores fail the approved tolerances.
D:Use relevant retrieval context, justified response limits and supported caching after verifying quality on evaluation examples. This is associated with prompt truncation strategy increases generated-token spend, not the essential evidence-and-fix pair for evaluation workflow fails to block degraded agent deployment. First examine: Inspect evaluation workflow outputs, release conditions, metric thresholds, CI job failure behavior and promotion records. Then correct: Make required quality/safety criteria explicit CI gates and stop deployment when evaluated scores fail the approved tolerances.
Question 719
During a Microsoft Foundry generative AI deployment review, the following condition occurs: A document-assistant project exists in Foundry but cannot infer because its configured model deployment name is from a different project and unavailable to the target identity. The deployment must retain governed identities, endpoint contracts and auditable release history. Which targeted Azure AI operations change best addresses the documented problem?
- Grant the narrowly scoped search data access role to the correct agent identity and use identity-based connection configuration.
- Define a versioned structured response instruction and supported JSON schema contract with parser-based validation.
- Separate governed development and production environments with distinct project configuration, least-privilege credentials and deployment gates.
- Bind the application to the reviewed compatible model deployment in the correct Foundry project/resource and validate the connection.
Explanation
Correct Answer
D:Bind the application to the reviewed compatible model deployment in the correct Foundry project/resource and validate the connection. This targets foundry project links to wrong deployed model resource. A project workspace and a model deployment are distinct resources; using a logical model name without the correct deployment binding is insufficient. Verify the failure through: Compare project connection configuration, model deployment resource ID, endpoint URL, deployment name and invocation authorization. Validate success with: Test queries reach the intended deployed model with attributable project usage and without cross-project credential exposure. This matters here because a document-assistant project exists in Foundry but cannot infer because its configured model deployment name is from….
Incorrect Answers
A:Grant the narrowly scoped search data access role to the correct agent identity and use identity-based connection configuration. This could address agent runtime managed identity cannot read retrieval data, but it does not solve the specified foundry project links to wrong deployed model resource failure. The matching remediation is: Bind the application to the reviewed compatible model deployment in the correct Foundry project/resource and validate the connection.
B:Define a versioned structured response instruction and supported JSON schema contract with parser-based validation. This could address prompt omits explicit structured answer schema, but it does not solve the specified foundry project links to wrong deployed model resource failure. The matching remediation is: Bind the application to the reviewed compatible model deployment in the correct Foundry project/resource and validate the connection.
C:Separate governed development and production environments with distinct project configuration, least-privilege credentials and deployment gates. This could address foundry environment mixes development and production agent resources, but it does not solve the specified foundry project links to wrong deployed model resource failure. The matching remediation is: Bind the application to the reviewed compatible model deployment in the correct Foundry project/resource and validate the connection.
Question 720
During a production Azure Machine Learning model lifecycle incident, the following condition occurs: An AutoML classification search reports near-perfect accuracy because the supplied training dataset includes a field recorded only after the real customer outcome is known. The operator must safeguard training reproducibility and serving continuity. Which investigation best verifies the cause before changing production configuration?
- Inspect class prevalence, AutoML primary metric, confusion matrix, recall and precision of candidate runs and business error costs.
- Inspect registered model metadata, online feature lookup mapping, feature timestamps, key names and schema contract from training.
- Review AutoML featurization, data collection timestamps, target leakage candidates, split strategy and training-versus-production feature availability.
- Inspect component reuse settings, input asset version, code and environment fingerprints, effective parameters and cached job lineage.
Explanation
Correct Answer
C:Review AutoML featurization, data collection timestamps, target leakage candidates, split strategy and training-versus-production feature availability. This is the most relevant evidence for automated ml experiment uses leakage feature. Automated model search cannot correct information leakage caused by improperly constructed input data and validation boundaries. The matching controlled remediation is: Remove target-leakage columns and rerun the AutoML search with an appropriate time-aware validation split and realistic available features.
Incorrect Answers
A:Inspect class prevalence, AutoML primary metric, confusion matrix, recall and precision of candidate runs and business error costs. This would investigate automl primary metric does not match rare-event goal rather than the documented automated ml experiment uses leakage feature issue. The relevant investigation is: Review AutoML featurization, data collection timestamps, target leakage candidates, split strategy and training-versus-production feature availability.
B:Inspect registered model metadata, online feature lookup mapping, feature timestamps, key names and schema contract from training. This would investigate feature retrieval specification missing from registered model rather than the documented automated ml experiment uses leakage feature issue. The relevant investigation is: Review AutoML featurization, data collection timestamps, target leakage candidates, split strategy and training-versus-production feature availability.
D:Inspect component reuse settings, input asset version, code and environment fingerprints, effective parameters and cached job lineage. This would investigate pipeline cache reuses stale transform artifact rather than the documented automated ml experiment uses leakage feature issue. The relevant investigation is: Review AutoML featurization, data collection timestamps, target leakage candidates, split strategy and training-versus-production feature availability.