Google Professional Machine Learning Engineer Question 324
Select 3Google Cloud PlatformYou are building a recommendation system for an e-commerce platform using Vertex AI. The system requires real-time feature data, such as user browsing history and product metadata, to generate predictions. You decide to use Vertex AI Feature Store. Which of the following steps are necessary to ensure that the feature store is properly set up for real-time serving?
- A
Create an online store for low-latency serving.
- B
Configure the feature store to use a batch export pipeline for feature data.
- C
Ingest feature data into both the online and offline stores.
- D
Define an entity type that represents the unique identifiers for your features.
- E
Set up a BigQuery table to directly serve features in real-time.
Show answer and explanation
Correct answers: A, C, D
Explanation
To use Vertex AI Feature Store for real-time serving, you need an online store to enable low-latency access to feature data. Feature data must be ingested into both online and offline stores to support different use cases. Additionally, defining entity types helps to uniquely identify and organize features. Batch pipelines and BigQuery are better suited for offline analytics rather than real-time serving.
- A. Correct.
Correct: An online store is required for low-latency serving, which is essential for real-time predictions.
- B. Incorrect.
Incorrect: A batch export pipeline is used primarily for offline analytics, not for real-time serving.
- C. Correct.
Correct: Feature data must be ingested into both the online and offline stores to support both real-time and batch use cases.
- D. Correct.
Correct: Defining an entity type is essential to organize and query features based on unique identifiers.
- E. Incorrect.
Incorrect: While BigQuery is useful for analytics, it is not designed for low-latency real-time serving. Vertex AI Feature Store’s online store should be used instead.