Google Vertex AI Cheat Sheet

Updated 2026-05-21

Next Topic: Graph Neural Networks (GNNs) Cheat Sheet

Google Vertex AI is Google Cloud's unified, fully managed MLOps platform for building, deploying, and managing machine learning and generative AI workloads at any scale. It covers the complete lifecycle — from interactive notebook development and dataset management through custom training, AutoML, and model serving to feature stores, pipeline orchestration, and production monitoring. The key mental model is that Vertex AI is not a single tool but a tightly integrated suite: nearly every resource (datasets, models, endpoints, experiments) lives in a Model Registry or dedicated store that pipelines, notebooks, and monitoring jobs reference by ID, making end-to-end reproducibility a first-class concern rather than an afterthought.

What This Cheat Sheet Covers

This topic spans 15 focused tables and 96 indexed concepts. Below is a complete table-by-table outline of this topic, spanning foundational concepts through advanced details.

Table 1: Notebook EnvironmentsTable 2: Managed DatasetsTable 3: AutoML TrainingTable 4: Custom TrainingTable 5: Vertex AI Pipelines (KFP v2)Table 6: Model Registry and VersioningTable 7: Online and Batch PredictionTable 8: Vertex AI Feature StoreTable 9: Model MonitoringTable 10: Vertex AI Vector SearchTable 11: Vertex AI Experiments and TensorBoardTable 12: Vertex AI Vizier (Hyperparameter Tuning)Table 13: Vertex AI Agent Builder and Agent EngineTable 14: Vertex AI Explainable AITable 15: Generative AI Training and Fine-Tuning

Table 1: Notebook Environments

Vertex AI offers two distinct interactive development environments, each targeting different use-case patterns and team needs — choosing the right one early prevents painful migrations later.

Type	Example	Description
Vertex AI Workbench Instance	`gcloud workbench instances create my-instance --machine-type=n1-standard-4 --location=us-central1-a`	JupyterLab on a Compute Engine VM; supports GPUs, custom conda environments, GitHub sync, and idle shutdown.
Colab Enterprise	Share notebook → set IAM → collaborate in real-time in browser	Managed, serverless notebook environment integrated with Vertex AI and BigQuery; emphasizes collaboration and Gemini AI-assisted coding.
Idle Shutdown	`--idle-shutdown-timeout=10800` (3 h)	Automatically stops (not terminates) the VM after the configured period of kernel inactivity; preserves disk but halts CPU/GPU billing.

Table 1: Notebook Environments

Vertex AI offers two distinct interactive development environments, each targeting different use-case patterns and team needs — choosing the right one early prevents painful migrations later.

Type	Example	Description
Vertex AI Workbench Instance	`gcloud workbench instances create my-instance --machine-type=n1-standard-4 --location=us-central1-a`	JupyterLab on a Compute Engine VM; supports GPUs, custom conda environments, GitHub sync, and idle shutdown.
Colab Enterprise	Share notebook → set IAM → collaborate in real-time in browser	Managed, serverless notebook environment integrated with Vertex AI and BigQuery; emphasizes collaboration and Gemini AI-assisted coding.
Idle Shutdown	`--idle-shutdown-timeout=10800` (3 h)	Automatically stops (not terminates) the VM after the configured period of kernel inactivity; preserves disk but halts CPU/GPU billing.