Skip to main content
Axalon Training
AI-104Axalon Original

Build AI Agents That Actually Work in Production

Master the complete Google Cloud agent development lifecycle — from ADK to AgentOps

Taught by Dan King, 4x Google Trainer of the Year

2 daysinstructor-ledIntermediate

Public class policy: Classes run with a minimum of 6 participants. If minimum enrollment isn't reached, you'll be notified 7 days before with options to transfer or receive a full refund.

From Zero to Deployed

Build and deploy working AI agents in 2 days. Go from foundational concepts to production deployment on Cloud Run and Vertex AI Agent Engine.

Beyond Prototypes

Learn what separates demo agents from production systems. The AgentOps framework gives you evaluation, monitoring, and CI/CD from day one.

Hands-On Every Module

Every module includes practical labs. Build LlmAgents, workflow agents, RAG pipelines, MCP integrations, and deploy them all.

Career-Ready Skills

Agent development is the fastest-growing skill in cloud engineering. This workshop gives you practical experience with the tools employers want.

Overview

A comprehensive 2-day workshop covering the complete lifecycle of building production-grade AI agents on Google Cloud. Starting with foundational concepts — models, tools, orchestration patterns like ReAct, and grounding techniques (RAG, GraphRAG, Agentic RAG) — participants progress to hands-on development with the Agent Development Kit (ADK). Learn to build LLM agents, workflow agents (Sequential, Parallel, Loop), implement tool ecosystems via MCP and A2A protocols, and deploy to Cloud Run, Vertex AI Agent Engine, and GKE. Day 2 focuses on production readiness through the AgentOps framework: 4-layer evaluation methodology, automated CI/CD with Agent Starter Pack, observability, and responsible AI practices including security guardrails and compliance.

What You'll Learn

  • Understand the core components of AI agents: models, tools, orchestration, and grounding
  • Build production-ready agents using Google Agent Development Kit (ADK)
  • Implement grounding techniques including RAG, GraphRAG, and Agentic RAG
  • Deploy agents to Cloud Run, Vertex AI Agent Engine, and GKE
  • Apply the AgentOps framework for evaluation, monitoring, and reliability
  • Design multi-agent systems using MCP and A2A protocols
  • Implement responsible AI practices and security safeguards

Who Should Attend

Cloud engineers, developers, and technical founders building AI agent systems on Google Cloud

Products Covered

Agent Development Kit (ADK)Vertex AIVertex AI Agent EngineCloud RunModel Context Protocol (MCP)Agent2Agent (A2A) ProtocolGeminiAgent Starter PackBigQueryFirestoreCloud Storage

Course Modules

1

Core Concepts of AI Agents

Topics

  • What is an AI agent
  • Three paths: build, use, partner
  • Google Cloud agent ecosystem overview
  • ADK vs Agentspace vs partner agents
  • MCP and A2A interoperability foundations

Activities

Group discussion: agentic vs traditional automationExplore Google Cloud agent ecosystem diagram
2

Key Components: Models, Tools & Data Architecture

Topics

  • Model selection: Flash-Lite vs Flash vs Pro
  • Model tuning and fine-tuning
  • Tool types and function calling
  • Data architecture: long-term, working, transactional memory
  • Google Cloud data services mapping

Activities

Lab: Compare Gemini model tiers on agent tasksDesign a data architecture for a sample agent
3

Orchestration & Grounding

Topics

  • ReAct framework: Reason, Act, Observe loop
  • RAG fundamentals and vector databases
  • GraphRAG: knowledge graph grounding
  • Agentic RAG: dynamic reasoning and retrieval
  • Vertex AI Search and RAG Engine
  • Grounding with Google Search

Activities

Lab: Build a RAG pipeline with Vertex AI SearchLab: Implement grounding with Google Search
4

Building with Agent Development Kit (ADK)

Topics

  • ADK agent types: LlmAgent, workflow agents, custom agents
  • SequentialAgent, ParallelAgent, LoopAgent patterns
  • Tool design: function signature, docstrings, return schema
  • FunctionTool and LongRunningFunctionTool
  • Agent-as-a-tool delegation pattern
  • Toolsets and BigQueryToolset

Activities

Lab: Build a multi-step LlmAgent with toolsLab: Create a SequentialAgent pipelineLab: Implement parallel data retrieval
5

MCP, A2A & Multi-Agent Systems

Topics

  • MCP: consuming and exposing tools
  • MCP Toolbox for Databases
  • A2A protocol: agent cards, task-oriented architecture
  • RemoteA2aAgent for distributed systems
  • Multi-agent composition patterns

Activities

Lab: Connect an ADK agent to MCP toolsLab: Build two agents that communicate via A2A
6

Deployment Patterns & Runtime

Topics

  • ADK deployment: adk api_server and containerization
  • Vertex AI Agent Engine: managed deployment
  • Cloud Run: serverless agent hosting
  • GKE: Kubernetes-based deployment
  • Memory Bank and Example Store
  • Choosing the right runtime

Activities

Lab: Deploy an ADK agent to Cloud RunLab: Deploy to Vertex AI Agent EngineDiscussion: runtime decision matrix
7

AgentOps: Evaluation & Observability

Topics

  • AgentOps methodology overview
  • Layer 1: Component-level testing
  • Layer 2: Trajectory evaluation
  • Layer 3: Outcome evaluation
  • Layer 4: System-level monitoring
  • Agent Starter Pack: IaC, CI/CD, observability
  • OpenTelemetry and Cloud Trace integration

Activities

Lab: Set up Agent Starter Pack projectLab: Write trajectory evaluation testsLab: Configure observability dashboards
8

Responsible AI & Security

Topics

  • Responsible AI principles for agents
  • Input/output guardrails and content filtering
  • Secure infrastructure with Terraform and IAM
  • Audit trails and compliance with BigQuery logging
  • Google Secure AI Framework (SAIF)
  • Defense-in-depth strategy

Activities

Lab: Implement prompt injection guardrailsDiscussion: SAIF risk assessment for your agent

Get This Training

No public classes currently scheduled. Express interest below or request private training.

Request Private Session

Course Details

Course Code
AI-104
Duration
2 days
Format
instructor-led
Level
Intermediate

What Makes This Workshop Different

Complete ADK Mastery

Build with every ADK agent type: LlmAgent for reasoning, SequentialAgent, ParallelAgent, and LoopAgent for workflows, and CustomAgent for unique requirements.

Real Deployment Skills

Deploy containerised agents to Cloud Run and Vertex AI Agent Engine. Learn the runtime decision matrix for choosing the right platform.

Grounding & RAG

Build RAG pipelines with Vertex AI Search and vector databases. Progress from basic RAG to GraphRAG and Agentic RAG patterns.

Production Evaluation

Go beyond vibe-testing with the 4-layer evaluation framework. Write trajectory tests, outcome evaluations, and set up system monitoring.

Multi-Agent Systems

Connect agents to external tools via MCP and build agent-to-agent communication with the A2A protocol.

Agent Starter Pack

Bootstrap production projects with Google official templates. Pre-configured Terraform, Cloud Build CI/CD, and evaluation pipelines.

“I had been tinkering with ADK for weeks but could not figure out how to get my agent into production properly. This workshop connected all the dots — deployment, evaluation, monitoring. My agent was running on Cloud Run by the end of day 2.”

Priya Sharma

Cloud Engineer, Technology Consultancy

Frequently Asked Questions

What do I need to know before attending?
You should be comfortable with Python and have basic familiarity with Google Cloud (console navigation, basic services). Prior AI/ML experience is helpful but not required — we cover foundational concepts before building.
Will I build a working agent during the workshop?
Yes. You will build multiple agents across the 2 days, from simple LlmAgents to multi-agent systems with MCP tools and A2A communication. You will deploy at least one agent to Cloud Run.
What tools and platforms will I use?
ADK (Python), Vertex AI Agent Engine, Cloud Run, Vertex AI Search, BigQuery, Firestore, Cloud Storage, Agent Starter Pack, and Cloud Trace. All labs run on Google Cloud with provided project access.
How is this different from the official Google Cloud training?
This workshop focuses specifically on production readiness — deployment, evaluation, AgentOps, and security. It is built by a Google Cloud Authorised Training Partner and goes deeper into hands-on development than standard courses.
Is this relevant if I am already building with ADK?
If you have built prototypes but not deployed to production, the AgentOps, deployment, and evaluation modules will be highly valuable. The workshop covers the gap between working demo and production system.
Do I get any certification or credential?
You receive an Axalon Training certificate of completion. The skills covered align with Google Cloud Professional Machine Learning Engineer and Cloud Developer certifications.
What format is the workshop?
Instructor-led with hands-on labs in every module. Approximately 60% hands-on lab time, 40% instruction and discussion. Available virtually or in-person.

Build Production AI Agents Your Team Can Rely On

From prototype to production — give your engineering team the complete Google Cloud agent development toolkit

Get Team QuoteContact Us
Free