Google Cloud Professional Agentic Architect Course

This course contains the use of artificial intelligence.

Prepare for the Google Cloud Professional Agentic Architect certification and learn to design agents you would trust in production.

Building an agent demo takes an afternoon. Running agents that serve real customers is a different job. Those agents need the right model, fresh and permission-aware data, tools that are safe to call, an evaluation suite that catches regressions, a deployment that scales, and security controls that hold even when a prompt is manipulated. The Google Cloud Professional Agentic Architect certification tests exactly those decisions, and this course teaches them.

The course follows the official exam guide from start to finish. You'll learn what each Google Cloud agent service does, and more importantly, when it is the right choice and when it is not. You'll follow one realistic company, Tidewater Mutual, a regional insurer, as it ships a low-code policy agent before hurricane season, modernizes its claims code with coding agents in a sandbox, builds and evaluates a multi-agent claims-intake system, deploys it to production, and handles a near-miss where an agent almost approves a payout it should not have.

What makes this course different

• Built around the official exam guide. Every section maps to one of the five exam domains, and the course spends its time in proportion to the domain weights. Developing custom agents, the largest domain, gets the most time.

• Current product names. Google renamed several services recently. The course uses the current names, such as Agent Runtime, Agent Search, and Gemini Enterprise, and points out the older names you'll still see in tutorials.

• Code and console walkthroughs. Where the certification expects hands-on skill, you'll see the actual code and configuration: ADK agents, callbacks, evalsets, deployments, revisions and traffic splits, Agent Gateway policies, and more.

• Exam-style reasoning throughout. Practice-scenario lectures in every domain show how to find the goal, the constraint, and the qualifier, and how to eliminate answers that technically work but don't fit.

• Practice included. A quiz after every section and a full-length, 80-question practice exam with an explanation for every option.

• About 5.4 hours of focused video (113 lectures), in short lectures with clean diagrams, rendered code, and step-by-step builds.

What you'll learn

Section 1: Welcome and orientation The two-part certification (a multiple-choice exam, then hands-on labs), the five domains and their weights, the Google Cloud agent stack in one map, and a study plan. You'll also meet Tidewater Mutual.

Section 2: Domain 1, Building agents using low-code tools - When low-code is the right call, and when it is not - Gemini Enterprise, Agent Designer, and the difference between chat agents and workflows - State-based conversations in CX Agent Studio: pages, transition routes, and event handlers - System instructions and prompt templates, including few-shot examples - Connecting enterprise data with connectors, data stores, and Agent Search - Multimodal data, and bringing custom agents into the Gemini Enterprise app

Section 3: Domain 2, Using coding agents for application development - What coding agents change for an architect, and the Antigravity app, IDE, CLI, and SDK - Giving coding agents tools with MCP servers - Secure sandboxes with GKE Agent Sandbox and Cloud Workstations - Refactoring, runtime optimization, and vulnerability patching - Customization with rules, skills, hooks, plugins, and subagents - Agents CLI for scaffolding, evaluating, deploying, and publishing agents

Section 4: Domain 3, Developing custom agents - The agent loop, design patterns, and choosing models from Model Garden - ADK fundamentals: tools, tool design, callbacks, sessions, state, and context engineering - Long-term memory with Memory Bank - RAG fundamentals, and choosing between RAG Engine, Vector Search, and Agent Search - Hybrid search, filtering, and reranking, plus structured data in BigQuery, Cloud SQL, Firestore, and Memorystore - Agent Identity, Agent Registry, and Google Cloud MCP servers - MCP versus A2A, agent cards, and hosting A2A agents - Workflow agents, LLM-driven delegation, graph workflows, and multi-agent systems in production

Section 5: Domain 4, Evaluating and deploying agentic workflows - Response and trajectory evaluation, test sets, evalsets, and autoraters - Judging retrieval and answer quality, continuous evaluation, and online monitors - User simulation and evaluating multi-agent systems - Choosing between Agent Runtime, Cloud Run, and GKE - Deployment, scaling, cold starts, concurrency, revisions, and canary rollouts - Observability with Cloud Trace and Cloud Logging, and troubleshooting drift, latency, and loops - Red-teaming, performance, reliability, and cost control

Section 6: Domain 5, Securing and governing agentic workflows - The agent threat model, including indirect prompt injection - OAuth 2.0, Agent Identity, and principal access boundary policies - Agent Gateway, Agent Registry, and Model Armor - Guardrails and human-in-the-loop approvals - Identity propagation, Sensitive Data Protection, private networking, and VPC Service Controls

Section 7: Labs, exam preparation, and review Six lab walkthroughs that rehearse the tasks the hands-on labs expect, an exam strategy for architecture scenarios, the ten most common wrong answers, mixed practice scenarios, and rapid reviews of every domain.

Practice that mirrors the exam

Each section ends with a short quiz. The course closes with a full-length, 80-question practice exam sized to the beta exam, following the official domain weighting and mixing multiple-choice and multiple-select questions. Every question explains each option, teaches the underlying concept, names the trap it tests, and links to the official Google Cloud documentation.

How to use this course to pass

1. Watch the sections in order. Tidewater Mutual's story builds from section to section.

2. Take each section quiz right away, and rewatch the related lectures if you score below 75%.

3. Pause during practice-scenario lectures, choose your answer, then listen to the reasoning.

4. Practice hands-on in your own Google Cloud project. The certification includes hands-on labs, and watching videos cannot replace doing the work. The lab walkthroughs show you exactly what to practice.

5. Take the full practice exam under timed conditions in your final week, then read every explanation.

Who this course is for

This course is designed for cloud architects, engineers, and developers preparing for the Professional Agentic Architect certification. It also suits solution architects and technical leads choosing agent platforms, and platform, security, and DevOps engineers who run and govern agents on Google Cloud. Familiarity with Google Cloud basics and some ability to read Python will help.

How the course is built

Lectures are short and focused, typically two to five minutes, so you can study in small blocks. Concept lectures explain what a service does and when to choose it. Worked examples apply those choices to Tidewater Mutual step by step. Lab walkthroughs show the code, commands, and console settings for a task, and practice-scenario lectures give you exam-style questions with the reasoning behind each answer. Many core lectures end with a deeper look at common traps: the answers that sound right but miss a constraint, such as relying on a prompt for a rule that must always hold, or picking a custom build when a managed service meets every requirement. All content, examples, and questions are original and checked against the current Google Cloud documentation.

About the exam

The certification has two parts: a proctored multiple-choice exam that tests design decisions and architectural standards, and hands-on labs in Google Skills that you're invited to after passing the exam. At the time of recording, the exam is in beta, with about 80 questions in three hours. Exam details can change at general availability, so confirm the current format, length, and price on the official Google Cloud certification page before you book.

Enroll now and build the judgment to pass the Professional Agentic Architect exam and design agents that hold up in production.

This course is not affiliated with, endorsed by, or sponsored by Google LLC. Google Cloud and related marks are trademarks of Google LLC.