Databricks Certified Machine Learning Professional Practice Test 40
Question 390
A production ML review contains this 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. What production-safe change best matches the evidence?
- Route availability errors to serving reliability remediation, not to automatic training.
- Maintain a model-version-linked training baseline alongside rolling comparisons.
- Require approved aggregation, privacy safeguards and sample sufficiency for sensitive slices.
- Configure a documented threshold with the responsible on-call team and a tested delivery route.
Explanation
Correct Answer
B:This corrects the relevant failure boundary instead of merely hiding the symptom. Maintain a model-version-linked training baseline alongside rolling comparisons. The issue matters because the live population drifts far from model training even though daily changes remain modest.
Incorrect Answers
A:That corrective action is directed at a different drift detection and lakehouse monitoring situation: An endpoint's error rate spikes while data profiles look normal, and responders attempt to retrain the model instead of checking serving infrastructure. The relevant corrective action here is maintain a model-version-linked training baseline alongside rolling comparisons.
C:That corrective action is directed at a different drift detection and lakehouse monitoring situation: A demographic fairness review uses a tiny sensitive subgroup with three labeled examples and reports its observed error as definitive. The relevant corrective action here is maintain a model-version-linked training baseline alongside rolling comparisons.
D:That corrective action 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. The relevant corrective action here is maintain a model-version-linked training baseline alongside rolling comparisons.
Question 391
Review this production model incident: A training notebook reports F1 for several models but only logs the final model's parameters in MLflow, leaving metrics and hyperparameters disconnected. Which failure risk follows most closely from the unresolved behavior?
- One customer's features can contaminate another customer's live score.
- Trial histories become detached from the actual optimization process and candidate provenance is lost.
- The deployed artifact cannot be tied to the training configuration that earned the winning metric.
- The apparent best F1 cannot be reproduced because its hyperparameters were never captured.
Explanation
Correct Answer
D:This consequence follows from the specific unresolved condition. The apparent best F1 cannot be reproduced because its hyperparameters were never captured. The causal constraint is: MLflow run-level logging should attach parameters and evaluated metrics to the corresponding candidate run.
Incorrect Answers
A:That direct consequence is directed at a different advanced mlflow usage situation: A fraud PyFunc uses a global mutable feature cache initialized by the first request and mixes data between subsequent users. The more immediate risk here is the apparent best F1 cannot be reproduced because its hyperparameters were never captured.
B:That direct consequence is directed at a different advanced mlflow usage situation: An experiment tunes 40 candidates but logs each trial as an unrelated root MLflow run, so reviewers cannot reconstruct which search produced a winner. The more immediate risk here is the apparent best F1 cannot be reproduced because its hyperparameters were never captured.
C:That direct consequence is directed at a different advanced mlflow usage situation: A data scientist selects the lowest validation loss from nested runs but reports only the parent run ID as the model-producing run. The more immediate risk here is the apparent best F1 cannot be reproduced because its hyperparameters were never captured.
Question 392
Review this production model incident: A training team averages gradients from multiple workers but lets each worker use a different untracked version of the model. Which mitigation resolves the defect rather than concealing it?
- Bind the selected model version to each group and join predictions on a unique provenance key.
- Retain SparkML for its supported distributed estimator or migrate to a validated Ray-compatible training algorithm.
- Pin shared validation protocol and data snapshot for the compared Optuna trials.
- Use framework-managed synchronized updates and reproducible checkpoints for the parallel training loop.
Explanation
Correct Answer
D:This corrects the relevant failure boundary instead of merely hiding the symptom. Use framework-managed synchronized updates and reproducible checkpoints for the parallel training loop. The issue matters because divergent stale weights compromise convergence and make failures difficult to reproduce.
Incorrect Answers
A:That corrective action is directed at a different scaling and tuning situation: A batch scoring pipeline trains one per-region model and then joins model predictions back using region alone, duplicating rows because each region contains multiple versioned models. The relevant corrective action here is use framework-managed synchronized updates and reproducible checkpoints for the parallel training loop.
B:That corrective action is directed at a different scaling and tuning situation: A team switches a SparkML training stage to Ray without verifying that its chosen estimator supports Ray's distributed fit contracts. The relevant corrective action here is use framework-managed synchronized updates and reproducible checkpoints for the parallel training loop.
C:That corrective action is directed at a different scaling and tuning situation: A hyperparameter team compares two Optuna studies with different train/validation splits and chooses the lower loss as though both metrics were directly comparable. The relevant corrective action here is use framework-managed synchronized updates and reproducible checkpoints for the parallel training loop.
Question 393
Consider the following fictional Databricks production case: A custom model endpoint meets mean response time of 75 milliseconds but P99 latency reaches 1.8 seconds during peak load. What consequence most specifically follows from leaving the issue unchanged?
- Routine weekend traffic launches unnecessary retraining and distracts engineers.
- A regression in v8 is masked by higher-volume predictions from v7.
- A small but critical share of customers experiences repeated slow predictions.
- The metric fails or silently excludes malformed observations, distorting the reported trend.
Explanation
Correct Answer
C:This consequence follows from the specific unresolved condition. A small but critical share of customers experiences repeated slow predictions. The causal constraint is: Tail latency can violate user experience even when mean latency appears healthy.
Incorrect Answers
A:That direct consequence 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. The more immediate risk here is a small but critical share of customers experiences repeated slow predictions.
B:That direct consequence 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. The more immediate risk here is a small but critical share of customers experiences repeated slow predictions.
D:That direct consequence is directed at a different drift detection and lakehouse monitoring situation: An ML owner adds a custom metric computed from the average of a string-typed currency column without validating conversion. The more immediate risk here is a small but critical share of customers experiences repeated slow predictions.
Question 394
Review this production model incident: A pipeline engineer fits a vectorizer after an estimator in the stage list and then tries to use the estimator's prediction as the vectorizer input. What evidence would best distinguish the actual fault from neighboring ML concerns?
- Compare train and validation event timestamps and detect any fold containing observations after its test period.
- Measure the local estimator runtime against cluster initialization and task scheduling for the same sample.
- Read the planned stage input/output dependencies and show the schema at the first failing stage.
- Compare absolute and squared residual distributions and inspect the high-error tail on holdout data.
Explanation
Correct Answer
C:This is the most discriminating check for the observed symptoms. Read the planned stage input/output dependencies and show the schema at the first failing stage. The underlying technical issue is that pipeline stages execute in declared sequence and each estimator expects columns already produced upstream.
Incorrect Answers
A:That diagnostic evidence is directed at a different using spark ml situation: A time-ordered customer-retention dataset includes events from the future, yet a developer randomly cross-validates across time and then claims a forward-looking forecast is unbiased. For the stated incident, the decisive observation is read the planned stage input/output dependencies and show the schema at the first failing stage.
B:That diagnostic evidence is directed at a different using spark ml situation: A laboratory runs a linear regression every quarter on only 280 clean observations and proposes a 48-worker Spark cluster to reduce a two-second local fit. For the stated incident, the decisive observation is read the planned stage input/output dependencies and show the schema at the first failing stage.
D:That diagnostic evidence is directed at a different using spark ml situation: A forecasting model has a few catastrophic numeric errors hidden by a low median absolute error and operations care about large misses. For the stated incident, the decisive observation is read the planned stage input/output dependencies and show the schema at the first failing stage.
Question 395
In an enterprise ML workflow, the reported issue is: A preprocessing function divides by a standard deviation and returns infinity for a constant numeric column. What technical distinction is essential to this investigation?
- Deterministic function tests should include degenerate data distributions that violate assumed mathematical denominators.
- Cross-component tests must validate shared time semantics and feature availability throughout the ML pipeline.
- Separating pure functions from orchestration can make unit tests faster while retaining notebook-driven workflows.
- Unit tests belong on controlled synthetic inputs before integrations, and production smoke tests do not replace them.
Explanation
Correct Answer
A:The technical constraint is specific to the reported failure. Deterministic function tests should include degenerate data distributions that violate assumed mathematical denominators. To test it, run a tiny constant-column fixture and inspect finite-value validation after transformation.
Incorrect Answers
B:That underlying principle is directed at a different validation testing situation: A feature-refresh task writes rows with event time in local timezone while the training join assumes UTC; unit tests of each component pass. The current incident instead turns on deterministic function tests should include degenerate data distributions that violate assumed mathematical denominators.
C:That underlying principle is directed at a different validation testing situation: A team hides all preprocessing logic inside one long notebook, making it hard to test independently without starting a Spark cluster. The current incident instead turns on deterministic function tests should include degenerate data distributions that violate assumed mathematical denominators.
D:That underlying principle is directed at a different validation testing situation: A developer changes a feature transform in a notebook and directly uses production data as the only test case. The current incident instead turns on deterministic function tests should include degenerate data distributions that violate assumed mathematical denominators.
Question 396
A production ML review contains this situation: A research team captures only daily batch inference totals with no row-level timestamps but insists on per-minute live inference drift analysis. Which check shows that this particular problem has been resolved?
- A known benign seasonal fixture stays below the revised threshold and a material anomaly exceeds it.
- New row-level timestamped records support the intended monitoring windows without interpolation guesses.
- A synthetic slow drift triggers the fixed-baseline comparison before the model exceeds tolerance.
- A retained timestamped fixture yields hour-separated metrics only under the appropriate profile.
Explanation
Correct Answer
B:This verification measures the required result rather than just process liveness. New row-level timestamped records support the intended monitoring windows without interpolation guesses. It confirms whether the targeted action succeeded: improve authorized logging and table design before configuring fine-grained inference time-series analysis.
Incorrect Answers
A:That acceptance test is directed at a different drift detection and lakehouse monitoring situation: A production alert threshold is hard-coded against last year's baseline, but seasonal traffic has shifted and the owner never reviewed the threshold. Here the actual success criterion is new row-level timestamped records support the intended monitoring windows without interpolation guesses.
C:That acceptance test is directed at a different drift detection and lakehouse monitoring situation: A streaming payments monitor always compares today with the previous hour, and a slow week-long distribution drift escapes because adjacent hours differ only slightly. Here the actual success criterion is new row-level timestamped records support the intended monitoring windows without interpolation guesses.
D:That acceptance test is directed at a different drift detection and lakehouse monitoring situation: 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. Here the actual success criterion is new row-level timestamped records support the intended monitoring windows without interpolation guesses.
Question 397
The engineering team encounters the following issue: A customer-risk feature needs a ratio of current debt to income, but online serving recomputes it differently from the training pipeline. What corrective action should the engineer prefer?
- Centralize the approved on-demand ratio transformation and its edge-case semantics.
- Apply supported event-time windows and watermark behavior to compute the intended rolling features.
- Enforce a supported lookback policy and explicit null handling for devices without recent data.
- Configure the supported synchronization process and freshness alerts for critical features.
Explanation
Correct Answer
A:This corrects the relevant failure boundary instead of merely hiding the symptom. Centralize the approved on-demand ratio transformation and its edge-case semantics. The issue matters because training-serving skew silently changes the meaning of risk thresholds.
Incorrect Answers
B:That corrective action is directed at a different advanced feature store concepts situation: A sensor feature pipeline computes rolling 30-minute averages by arrival time, although delayed events carry trustworthy event timestamps. The relevant corrective action here is centralize the approved on-demand ratio transformation and its edge-case semantics.
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 centralize the approved on-demand ratio transformation and its edge-case semantics.
D: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 centralize the approved on-demand ratio transformation and its edge-case semantics.
Question 398
Review this production model incident: A hyperparameter search operates on large Spark-managed feature tables but each candidate model is small and independent, and workers repeatedly convert the entire table for each Ray task. Which failure risk follows most closely from the unresolved behavior?
- Training stalls on memory pressure even while unrelated cluster nodes are underutilized.
- A trial's metrics and model artifacts overwrite evidence from another configuration.
- Divergent stale weights compromise convergence and make failures difficult to reproduce.
- Repeated cross-engine transfers dominate tuning time and may exhaust network capacity.
Explanation
Correct Answer
D:This consequence follows from the specific unresolved condition. Repeated cross-engine transfers dominate tuning time and may exhaust network capacity. The causal constraint is: Combining Spark feature preparation with Ray trial execution requires explicit materialization and data locality planning.
Incorrect Answers
A:That direct consequence is directed at a different scaling and tuning situation: A single-node model training run spends almost all time swapping a 40 GB feature matrix on a 16 GB VM; no distributed version exists. The more immediate risk here is repeated cross-engine transfers dominate tuning time and may exhaust network capacity.
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 repeated cross-engine transfers dominate tuning time and may exhaust network capacity.
C:That direct consequence is directed at a different scaling and tuning situation: A training team averages gradients from multiple workers but lets each worker use a different untracked version of the model. The more immediate risk here is repeated cross-engine transfers dominate tuning time and may exhaust network capacity.
Question 399
In an enterprise ML workflow, the reported issue is: A model serving service uses a broadly privileged shared token to read every online feature table when it needs only one approved feature set. What likely operational or model-quality harm does this issue create?
- One purchase increases the spending feature twice, affecting downstream risk scores.
- Credential compromise exposes unrelated customers' private feature data.
- Training labels duplicate and features from future measurements become attached to older events.
- External API variability causes widespread prediction timeouts at peak traffic.
Explanation
Correct Answer
B:This consequence follows from the specific unresolved condition. Credential compromise exposes unrelated customers' private feature data. The causal constraint is: Online feature access should use least-privilege identity and scoped permissions rather than universal shared credentials.
Incorrect Answers
A:That direct consequence is directed at a different advanced feature store concepts situation: A real-time feature job updates user purchase totals without checking whether upstream transaction events are delivered twice. The more immediate risk here is credential compromise exposes unrelated customers' private feature data.
C:That direct consequence 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 more immediate risk here is credential compromise exposes unrelated customers' private feature data.
D:That direct consequence is directed at a different advanced feature store concepts situation: A streaming recommender's request-time feature function calls a slow third-party API synchronously on every prediction. The more immediate risk here is credential compromise exposes unrelated customers' private feature data.
Question 400
Review this production model incident: A model serving endpoint is created manually in the UI and every staging deployment differs subtly from production because no deployment definition is saved. Which mitigation resolves the defect rather than concealing it?
- Define the supported model-serving and MLflow resources declaratively with validated environment targets.
- Apply approved workload isolation, compute policies, quotas and cost attribution for training versus serving.
- Move scheduled ML pipelines to governed service principals and approved compute policies.
- Deploy region-appropriate governed feature and serving assets with controlled cross-region access.
Explanation
Correct Answer
A:This corrects the relevant failure boundary instead of merely hiding the symptom. Define the supported model-serving and MLflow resources declaratively with validated environment targets. The issue matters because manual drift creates unreproducible endpoint permissions or traffic settings.
Incorrect Answers
B:That corrective action is directed at a different environment architectures situation: A production deployment scales to 30 teams but all jobs share one compute pool with no capacity limits or cost attribution. The relevant corrective action here is define the supported model-serving and MLflow resources declaratively with validated environment targets.
C:That corrective action is directed at a different environment architectures situation: A distributed retraining workflow uses ephemeral personal compute and fails when the original engineer's account is disabled. The relevant corrective action here is define the supported model-serving and MLflow resources declaratively with validated environment targets.
D:That corrective action is directed at a different environment architectures situation: A multilingual prediction service must handle regional data residency, but the team copies a shared feature cache across regions without governance. The relevant corrective action here is define the supported model-serving and MLflow resources declaratively with validated environment targets.