
Instead of combining separate tools across different stages of a project, teams on Google Cloud can train, host, and serve machine learning models within a single integrated environment using Gemini Enterprise Agent Platform (formerly Vertex AI). As Google Cloud's end-to-end platform for machine learning operations and AI agents, it is a central service on the Professional Machine Learning Engineer exam. In documentation and exam scenarios, the service is often shortened to Agent Platform, and both names refer to the same product.
Teams can work with Agent Platform through the Google Cloud console or programmatically through the Agent Platform SDK, the Agent Platform API, and the gcloud CLI. Programmatic access is what enables automated workflows without manual console steps. When Google Cloud expanded the service into Gemini Enterprise Agent Platform, none of the existing model tooling was removed. The platform is organized into two primary sections. The Models section covers the classic machine learning workflow of training, tuning, deploying, and monitoring models. The Agents section is the newer addition, focused on building and running AI agents that can reason and use tools. For the Professional Machine Learning Engineer exam, we would keep the relative weight of those two sections in mind: both are worth knowing, though Agents currently carries much less weight than Models.
Historically, machine learning teams ran into four recurring operational issues when managing workflows across disconnected tooling. Agent Platform is designed to address these issues inside one environment.
The Models section brings together the established components for managing an end-to-end machine learning workflow without constantly switching tools or moving data and models across platforms. Not every feature in the Models section is tested on the exam, so exam study should center on the core components across data preparation, model development, model management, deployment, and monitoring.
The Agents section gathers the services needed to build an AI agent, ground it in organizational data, and connect it to external tools without leaving Agent Platform. Its components fall into three functional groups.
Discovery and development covers where agents are created and configured. Agent Garden provides a library of ready-made agent samples and templates that teams can browse and use as a starting point instead of designing from scratch. Agent Studio is the visual environment in the console for configuring an agent, writing instructions, attaching tools, and testing behavior interactively without writing much code. Agent Development Kit (ADK) is the code-first framework for the same task, giving developers structure to define an agent, its instructions, and the tools it is allowed to call when full control over behavior is required.
Grounding and retrieval connects an agent to an organization's own data. Because a model on its own only knows what it learned during training, company-specific information must be retrieved and supplied at request time. Three services handle this part of the workflow:
Tool integration allows an agent to take action outside Agent Platform through MCP servers, which are based on the Model Context Protocol. MCP servers offer a standard interface for exposing external tools and data sources to an agent. Through that consistent contract, an agent can call an API, query a system, or trigger an action instead of relying on a custom integration for every external tool.
When preparing for the Professional Machine Learning Engineer exam, start by treating Gemini Enterprise Agent Platform and Agent Platform as interchangeable names for the same unified service. Spend most of your time on the Models section, from managed datasets and Feature Store through Model Registry, prediction endpoints, and post-deployment monitoring. Alongside that foundation, know how the Agents section divides responsibilities across discovery and development, grounding and retrieval, and tool integration, paying attention to where Google Cloud offers a managed option versus a lower-level or code-first option for greater control.
Our Professional Machine Learning Engineer course covers Gemini Enterprise Agent Platform alongside custom training and model monitoring, with practice questions that drill these distinctions.