Cloud Digital Leader Google Cloud Products and Services Practice Question

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✓

  Vertex AI

  Why this is correct

  Vertex AI provides a unified managed platform that covers the full ML lifecycle: training with custom or pre-built containers on GPU/TPU, versioning in a Model Registry, and deployment to prediction endpoints with built-in autoscaling based on traffic. It eliminates the need to manually configure infrastructure, allows custom model serving for any framework, and offers MLOps capabilities like monitoring and drift detection. This makes it the recommended service for training and serving TensorFlow models.

• ✗

  AI Platform (legacy)

  Why it's wrong here

  AI Platform (legacy) originated as a separate ML platform and is now absorbed into Vertex AI as a legacy service. While it supports training and prediction, it lacks integration with Vertex AI's unified pipelines, feature store, and model monitoring, and Google is not investing in new features for it—only maintenance. The correct choice for new projects is Vertex AI because it provides a consistent interface for batch and online prediction, as well as a migration path from legacy AI Platform.

• ✗

  Cloud Functions

  Why it's wrong here

  Cloud Functions is an event-driven, serverless compute service designed for short-lived, single-purpose functions (e.g., webhooks, lightweight data processing). It does not support GPUs, imposes strict execution timeouts (typically up to 540 seconds), and is not built for long-running ML training or continuous HTTP model serving with autoscaling. Consequently, it cannot handle TensorFlow training or production-grade online inference where request patterns are variable and require model-specific infrastructure.

• ✗

  Compute Engine with pre-installed ML frameworks

  Why it's wrong here

  Compute Engine with pre-installed ML frameworks gives you raw VMs, which require you to manually manage GPU drivers, framework versions, and serving dependencies. To achieve autoscaling for prediction, you must build your own infrastructure: set up instance groups, health checks, load balancers, and metrics-based autoscaling policies—all of which Vertex AI provides out of the box. This operational overhead increases risk and maintenance burden, and it is not a purpose-built solution for the end-to-end ML workflow.

Go deeper

Related to this question