PMLE - Collaborating Within and Across Teams to Manage Data and Models - Section 2.1

Explore and preprocess data for ML across tabular, text, and image types, choosing the right tool for scale such as BigQuery, Dataflow, Apache Spark, and in-memory Python frameworks, consolidating features in the Agent Platform Feature Store, and protecting personally identifiable information.

Choose between Dataflow, Apache Spark, and in-memory Python frameworks based on data volume and type, and consolidate reusable features in the Agent Platform Feature Store to avoid training-serving inconsistency. Recognise which PII-handling techniques - such as tokenisation and data masking - are appropriate when preprocessing sensitive tabular, text, or image data.

DataflowApache SparkAgent Platform Feature StorePII handling

Practice question for this objective

Free sampleCollaborating Within and Across Teams to Manage Data and Modelsmedium

A fraud team has computed a customer's rolling 30-day transaction count in a batch job and now needs the identical value served at low latency during online inference, while guaranteeing that training data drawn from history matches what the model will see in production. Their platform offers the Agent Platform Feature Store. Which capability of the Feature Store most directly prevents training-serving skew for this feature?

  • AA managed offline store for point-in-time training reads and an online store serving the same feature definition at low latency. Correct
  • BAutomatic hyperparameter tuning that retrains the fraud model whenever the rolling-count feature distribution drifts beyond a threshold.
  • CColumn-level encryption that masks the transaction count so that training and serving both read an obfuscated value.
  • DA built-in cache that stores recent model predictions so repeated inference requests reuse the previous scored output.
Recognise that a feature store prevents training-serving skew via shared definitions across point-in-time offline and online stores. Training-serving skew arises when training and inference compute or read a feature differently; a feature store registers one definition served from an online store and read point-in-time from an offline store, so the same logic feeds both paths.

Why A is correct: The Feature Store pairs an offline store that supports point-in-time correct training lookups with an online store serving the same registered feature, so both paths read one definition and skew is avoided by construction.

Why B is wrong: Drift-triggered retraining is a real lifecycle practice and sounds relevant to changing features, but it addresses model staleness rather than the Feature Store's job of serving consistent feature values across training and inference.

Why C is wrong: Encryption and masking protect sensitive fields and could plausibly be confused with consistency controls, but obfuscating a value does nothing to keep training and serving reads aligned, which is the skew problem here.

Why D is wrong: Prediction caching can cut latency and seems to help serving, but it stores outputs rather than feature inputs and never touches how training reads historical features, so it cannot prevent skew.

See more PMLE practice questions with worked answers.

More in this domain

Back to all Collaborating Within and Across Teams to Manage Data and Models objectives, or the PMLE cert hub.

Examworthy is not affiliated with or endorsed by Google Cloud. Original, blueprint-aligned practice material only.