Databricks Certified Machine Learning Professional Practice Test 74
Question 730
A production ML review contains this situation: A feature refresh writes duplicate records for each entity and logical timestamp after retries, breaking unique lookups. Which statement best explains the failure mode?
- A time-series feature table needs a documented timestamp key and an as-of lookup to avoid multiplying labels or attaching future values.
- Idempotent feature pipeline writes need stable entity/time keys and safe retry semantics.
- On-demand feature definitions should apply consistent logic and missing/zero-denominator handling across contexts.
- Low-latency serving requires online feature access with a documented freshness and lookup contract.
Explanation
Correct Answer
B:The technical constraint is specific to the reported failure. Idempotent feature pipeline writes need stable entity/time keys and safe retry semantics. To test it, check entity/timestamp duplicate counts around job retries and writer transaction history.
Incorrect Answers
A:That underlying principle is directed at a different advanced feature store concepts situation: A training dataset combines a feature table with multiple timestamped rows per device using an equality join on device ID only. The current incident instead turns on idempotent feature pipeline writes need stable entity/time keys and safe retry semantics.
C:That underlying principle is directed at a different advanced feature store concepts situation: A customer-risk feature needs a ratio of current debt to income, but online serving recomputes it differently from the training pipeline. The current incident instead turns on idempotent feature pipeline writes need stable entity/time keys and safe retry semantics.
D:That underlying principle is directed at a different advanced feature store concepts situation: A pricing endpoint queries a large offline Delta feature table for every user request, and P95 response time reaches several seconds. The current incident instead turns on idempotent feature pipeline writes need stable entity/time keys and safe retry semantics.
Question 731
The engineering team encounters the following issue: An interrupted Ray hyperparameter sweep has an expensive completed-trial history but a new launch discards its experiment checkpoint. What direct consequence should the owner prioritize if no fix is made?
- Oversubscribed trials spend time waiting and can destabilize shared model training.
- A trial's metrics and model artifacts overwrite evidence from another configuration.
- A costly driver upgrade leaves worker-side backlog and deadline misses unchanged.
- Training compute is wasted and selected parameters are biased toward only the new subset of trials.
Explanation
Correct Answer
D:This consequence follows from the specific unresolved condition. Training compute is wasted and selected parameters are biased toward only the new subset of trials. The causal constraint is: Fault-tolerant distributed tuning should use supported experiment storage and restoration rather than recreating every trial.
Incorrect Answers
A:That direct consequence is directed at a different scaling and tuning situation: An operations model uses a SparkML estimator that itself distributes data, yet the team launches dozens of concurrent full-cluster fits on the same fixed small cluster. The more immediate risk here is training compute is wasted and selected parameters are biased toward only the new subset of trials.
B:That direct consequence is directed at a different scaling and tuning situation: An Optuna optimization reuses the same MLflow run ID for concurrent trials, causing overlapping parameter values and artifacts. The more immediate risk here is training compute is wasted and selected parameters are biased toward only the new subset of trials.
C:That direct consequence is directed at a different scaling and tuning situation: A feature-engineering job is CPU-bound on hundreds of independent daily partitions and a team upgrades only the driver to a much larger machine. The more immediate risk here is training compute is wasted and selected parameters are biased toward only the new subset of trials.
Question 732
Consider the following fictional Databricks production case: A real-time feature job updates user purchase totals without checking whether upstream transaction events are delivered twice. Which implementation is most appropriate for the stated incident?
- Centralize eligible feature definitions and reuse a controlled feature computation workflow where supported.
- Configure the supported synchronization process and freshness alerts for critical features.
- Enforce a supported lookback policy and explicit null handling for devices without recent data.
- Add a supported duplicate-safe event processing strategy and preserve source identifiers.
Explanation
Correct Answer
D:This corrects the relevant failure boundary instead of merely hiding the symptom. Add a supported duplicate-safe event processing strategy and preserve source identifiers. The issue matters because one purchase increases the spending feature twice, affecting downstream risk scores.
Incorrect Answers
A:That corrective action is directed at a different advanced feature store concepts situation: An ML pipeline computes training features in one notebook and duplicates slightly different SQL in the online pipeline for the same entities. The relevant corrective action here is add a supported duplicate-safe event processing strategy and preserve source identifiers.
B:That corrective action is directed at a different advanced feature store concepts situation: A team publishes an online feature table but neglects its refresh schedule while the predictive model assumes values no older than 15 minutes. The relevant corrective action here is add a supported duplicate-safe event processing strategy and preserve source identifiers.
C:That corrective action is directed at a different advanced feature store concepts situation: A sensor classifier silently reuses features from a year-old device after months without readings, despite a maximum allowed lookback of seven days. The relevant corrective action here is add a supported duplicate-safe event processing strategy and preserve source identifiers.
Question 733
The engineering team encounters the following issue: A retail profile is created on a static materialized daily inventory table; a team expects its initial snapshot to show each prior hour as a separate trend. Which experiment most directly demonstrates a successful fix?
- A retained timestamped fixture yields hour-separated metrics only under the appropriate profile.
- A safe synthetic threshold crossing reaches the correct team with the monitored metric and window.
- Weekend fixtures match appropriate seasonality while atypical weekend changes still alert.
- A synthetic request can be traced to a versioned, privacy-safe inference record and matching endpoint response.
Explanation
Correct Answer
A:This verification measures the required result rather than just process liveness. A retained timestamped fixture yields hour-separated metrics only under the appropriate profile. It confirms whether the targeted action succeeded: use an appropriate time-series event table/profile if hourly history is truly required, keeping snapshot monitoring for current state.
Incorrect Answers
B:That acceptance test is directed at a different drift detection and lakehouse monitoring situation: A fraud-feature drift monitor crosses its threshold but the alert has no assigned recipient and no incident ownership. Here the actual success criterion is a retained timestamped fixture yields hour-separated metrics only under the appropriate profile.
C:That acceptance test is directed at a different drift detection and lakehouse monitoring situation: An online feature monitor compares weekend shopping to weekday reference data and treats expected weekend behavior as a model failure. Here the actual success criterion is a retained timestamped fixture yields hour-separated metrics only under the appropriate profile.
D:That acceptance test is directed at a different drift detection and lakehouse monitoring situation: A serving endpoint appears healthy but nobody records model version, request identifier or prediction value, so a drift incident cannot be reconstructed. Here the actual success criterion is a retained timestamped fixture yields hour-separated metrics only under the appropriate profile.
Question 734
Consider the following fictional Databricks production case: A binary classifier ranks risky cases well but sends too many alerts at a default 0.5 decision threshold. Which observation most directly tests the suspected failure boundary?
- Compare request latency and concurrency service-level objectives with the batch schedule and observed turnaround.
- Inspect the scoring job for collect and driver memory growth versus distributed transform stages.
- Review predicted probabilities, a precision-recall threshold curve and analyst capacity at candidate cutoffs.
- Inspect its declared output column and confirm the downstream estimator consumes a vector rather than predicted labels.
Explanation
Correct Answer
C:This is the most discriminating check for the observed symptoms. Review predicted probabilities, a precision-recall threshold curve and analyst capacity at candidate cutoffs. The underlying technical issue is that ranking quality and operating threshold quality are distinct; precision and recall depend on the chosen cutoff.
Incorrect Answers
A:That diagnostic evidence is directed at a different using spark ml situation: An online pricing service requires a personalized response within 70 milliseconds per request but its team schedules a full overnight batch job as the only inference route. For the stated incident, the decisive observation is review predicted probabilities, a precision-recall threshold curve and analyst capacity at candidate cutoffs.
B:That diagnostic evidence is directed at a different using spark ml situation: A logistics model must score 60 million daily routes already in Delta but the scoring script collects each route into Python before calling predict. For the stated incident, the decisive observation is review predicted probabilities, a precision-recall threshold curve and analyst capacity at candidate cutoffs.
D:That diagnostic evidence is directed at a different using spark ml situation: A team wants to convert several numeric inputs into one features vector but treats VectorAssembler as a clustering Estimator that should output predictions. For the stated incident, the decisive observation is review predicted probabilities, a precision-recall threshold curve and analyst capacity at candidate cutoffs.
Question 735
Review this production model incident: A serving client calls the REST endpoint expecting one number but the custom PyFunc returns a structured list of prediction and confidence fields. Which failure risk follows most closely from the unresolved behavior?
- Recovery requires guessing traffic splits, permissions and served model versions during an outage.
- A malformed request crashes the entire serving worker or produces an untraceable score.
- Manual and automated updates race, creating unexpected model routing during production traffic.
- Clients throw parsing errors or accidentally interpret confidence as the prediction.
Explanation
Correct Answer
D:This consequence follows from the specific unresolved condition. Clients throw parsing errors or accidentally interpret confidence as the prediction. The causal constraint is: Client integrations should parse the documented model output contract instead of assuming a scalar response.
Incorrect Answers
A:That direct consequence is directed at a different custom model serving situation: A platform engineer creates a custom serving endpoint manually and assumes the exact settings can be recovered after an accidental deletion. The more immediate risk here is clients throw parsing errors or accidentally interpret confidence as the prediction.
B:That direct consequence is directed at a different custom model serving situation: A custom PyFunc's predict method assumes every request contains a numeric feature that is optional under the documented inference schema. The more immediate risk here is clients throw parsing errors or accidentally interpret confidence as the prediction.
C:That direct consequence is directed at a different custom model serving situation: A team edits a model endpoint through the UI while its CI pipeline concurrently applies an older configuration and reverts the intended traffic split. The more immediate risk here is clients throw parsing errors or accidentally interpret confidence as the prediction.
Question 736
In an enterprise ML workflow, the reported issue is: An online behavioral feature counter consumes a streaming source but does not retain checkpoint state after worker restarts. Before modifying the pipeline, which evidence should be collected first?
- Compare last available reading age with each prediction cutoff and its permitted freshness window.
- Inspect checkpoint paths and duplicate feature writes after a simulated restart.
- Inspect feature refresh run metadata, table versions and transformation code revisions.
- Compare offline and serving results for synthetic zero-income, missing and normal customer fixtures.
Explanation
Correct Answer
B:This is the most discriminating check for the observed symptoms. Inspect checkpoint paths and duplicate feature writes after a simulated restart. The underlying technical issue is that fault-tolerant real-time feature computation requires durable streaming progress and suitable sink idempotency.
Incorrect Answers
A:That diagnostic evidence is directed at a different advanced feature store concepts situation: A sensor classifier silently reuses features from a year-old device after months without readings, despite a maximum allowed lookback of seven days. For the stated incident, the decisive observation is inspect checkpoint paths and duplicate feature writes after a simulated restart.
C:That diagnostic evidence is directed at a different advanced feature store concepts situation: A team manually reruns a feature notebook each week before training; feature definitions change without an execution history or stable destination. For the stated incident, the decisive observation is inspect checkpoint paths and duplicate feature writes after a simulated restart.
D:That diagnostic evidence is directed at a different advanced feature store concepts situation: A customer-risk feature needs a ratio of current debt to income, but online serving recomputes it differently from the training pipeline. For the stated incident, the decisive observation is inspect checkpoint paths and duplicate feature writes after a simulated restart.
Question 737
In an enterprise ML workflow, the reported issue is: A notebook uses the MLflow Deployments SDK to send a column-major payload to a custom model endpoint expecting a different documented dataframe schema. What technical distinction is essential to this investigation?
- Serving inference payloads must conform to the endpoint and model's accepted request/response schema.
- Custom model input contracts should validate missing and malformed values before feature transformation.
- Model artifacts should not serialize secrets; serving identities should resolve credentials securely at runtime.
- Repeatable endpoint deployment should preserve a controlled configuration definition and registered model identity.
Explanation
Correct Answer
A:The technical constraint is specific to the reported failure. Serving inference payloads must conform to the endpoint and model's accepted request/response schema. To test it, compare serialized request JSON field names, batch dimensions and the model signature for a failing fixture.
Incorrect Answers
B:That underlying principle is directed at a different custom model serving situation: A custom PyFunc's predict method assumes every request contains a numeric feature that is optional under the documented inference schema. The current incident instead turns on serving inference payloads must conform to the endpoint and model's accepted request/response schema.
C:That underlying principle is directed at a different custom model serving situation: An engineer logs a customer-specific API token inside the pickled PythonModel object before registering it to Unity Catalog. The current incident instead turns on serving inference payloads must conform to the endpoint and model's accepted request/response schema.
D:That underlying principle is directed at a different custom model serving situation: A platform engineer creates a custom serving endpoint manually and assumes the exact settings can be recovered after an accidental deletion. The current incident instead turns on serving inference payloads must conform to the endpoint and model's accepted request/response schema.
Question 738
Consider the following fictional Databricks production case: A new recommendation model consumes three times as much memory and the team plans a canary based solely on prediction accuracy, ignoring serving capacity. Which observation most directly tests the suspected failure boundary?
- Replay stored valid requests from old and new clients against the candidate endpoint before shifting traffic.
- Compare replica memory, P95 latency and model quality for both versions under realistic canary load.
- Inspect endpoint state, individual served-entity readiness and model-load errors after the update.
- Compare effective serving traffic split, model identities and alias resolution in the live endpoint.
Explanation
Correct Answer
B:This is the most discriminating check for the observed symptoms. Compare replica memory, P95 latency and model quality for both versions under realistic canary load. The underlying technical issue is that rollout safety needs operational latency and memory gates alongside predictive quality metrics.
Incorrect Answers
A:That diagnostic evidence is directed at a different deployment strategies situation: A rollout updates the serving model but a renamed input column breaks requests from an older mobile client. For the stated incident, the decisive observation is compare replica memory, P95 latency and model quality for both versions under realistic canary load.
C:That diagnostic evidence is directed at a different deployment strategies situation: An endpoint update request succeeds but the newly configured model version is still loading; monitoring marks the release complete immediately. For the stated incident, the decisive observation is compare replica memory, P95 latency and model quality for both versions under realistic canary load.
D:That diagnostic evidence is directed at a different deployment strategies situation: During a controlled release, the candidate model's P95 latency triples and engineers want to change the registered model alias without adjusting endpoint routing. For the stated incident, the decisive observation is compare replica memory, P95 latency and model quality for both versions under realistic canary load.
Question 739
In an enterprise ML workflow, the reported issue is: A demographic fairness review uses a tiny sensitive subgroup with three labeled examples and reports its observed error as definitive. Which change addresses the real cause without undermining other requirements?
- Join labels to their originating predictions using stable IDs and a defined maturity window.
- Maintain a model-version-linked training baseline alongside rolling comparisons.
- Create a dedicated approved critical-region alert with suitable minimum volume and failure-rate conditions.
- Require approved aggregation, privacy safeguards and sample sufficiency for sensitive slices.
Explanation
Correct Answer
D:This corrects the relevant failure boundary instead of merely hiding the symptom. Require approved aggregation, privacy safeguards and sample sufficiency for sensitive slices. The issue matters because small-cell disclosure and unstable rates cause privacy risk and misleading fairness decisions.
Incorrect Answers
A:That corrective action is directed at a different drift detection and lakehouse monitoring situation: A model performance dashboard compares this week's labels to last week's predictions without joining by stable request ID or consistent outcome horizon. The relevant corrective action here is require approved aggregation, privacy safeguards and sample sufficiency for sensitive slices.
B:That corrective action is directed at a different drift detection and lakehouse monitoring situation: A model trained on last year's customer mix is monitored only against yesterday's already shifted cohort, so its original training-population mismatch is hidden. The relevant corrective action here is require approved aggregation, privacy safeguards and sample sufficiency for sensitive slices.
C:That corrective action is directed at a different drift detection and lakehouse monitoring situation: An alert query uses one global transaction-failure count and masks a complete outage in a small high-value region. The relevant corrective action here is require approved aggregation, privacy safeguards and sample sufficiency for sensitive slices.
Question 740
In an enterprise ML workflow, the reported issue is: A global model predicts late deliveries accurately overall, but all mistakes occur at one small coastal depot and the team never slices its results by site. Before modifying the pipeline, which evidence should be collected first?
- Inspect inference table version IDs and label-matched errors before and after traffic shift.
- Compare model errors and sample sizes by depot against the aggregate baseline.
- Compare per-replica resident memory and allocation limits before and after model promotion.
- Stratify historical samples by day type and examine whether the observed difference is expected seasonality.
Explanation
Correct Answer
B:This is the most discriminating check for the observed symptoms. Compare model errors and sample sizes by depot against the aggregate baseline. The underlying technical issue is that slicing on operationally meaningful locations can reveal concentrated failures hidden in the global metric.
Incorrect Answers
A:That diagnostic evidence is directed at a different drift detection and lakehouse monitoring situation: A scoring endpoint quietly moves from model v7 to v8 during a rollout but an inference trend aggregates both versions in one error curve. For the stated incident, the decisive observation is compare model errors and sample sizes by depot against the aggregate baseline.
C:That diagnostic evidence is directed at a different drift detection and lakehouse monitoring situation: A new model version uses twice the memory per replica, but capacity monitoring tracks only CPU and crash counts. For the stated incident, the decisive observation is compare model errors and sample sizes by depot against the aggregate baseline.
D:That diagnostic evidence is directed at a different drift detection and lakehouse monitoring situation: An online feature monitor compares weekend shopping to weekday reference data and treats expected weekend behavior as a model failure. For the stated incident, the decisive observation is compare model errors and sample sizes by depot against the aggregate baseline.