GTR = sesiune garantata

*Pretul nu include TVA si este valabil pentru un participant inscris la sesiuni publice sau la un curs in format SPVC sau WBT.
Inscrierea la o sesiune publica nu va garanteaza rezervarea unui loc. La scurt timp de la momentul inscrierii, veti fi contactat de un reprezentant Kluger pentru confirmarea rezervarii si a pretului final.
Disponibilitatea acestui curs in alte formate de livrare, sub un cod diferit, se poate verifica in portofoliu.
Daca doriti o sesiune privata, cu agenda standard sau adaptata nevoilor dvs., va rugam sa ne contactati.

Agentic AI with Google Cloud is a hands-on, one-day training introducing participants to Generative AI applications and AI agents using Gemini, Vertex AI, Google ADK, MCP, BigQuery, RAG, and Cloud Run.
The course follows the practical examples from a GitHub repository, progressing from a simple Gemini application to tool-using agents, an MCP-enabled analytics agent, a planner–executor–reporter workflow with dynamic routing, and deployment to Cloud Run.

By the end of the training, participants will be able to:

• Use Gemini through Vertex AI and Python.
• Build and test agents with Google ADK.
• Add Python tools and structured output.
• Connect an agent to BigQuery through an MCP server.
• Understand planner–executor–reporter orchestration.
• Use dynamic routing between tools, including a RAG tool.
• Deploy an ADK application to Cloud Run.

Cursul se adresează profesioniștilor IT care doresc să proiecteze, dezvolte și implementeze aplicații Agentic AI pe Google Cloud

• • Basic Python knowledge.
  • Familiarity with APIs and command-line tools.
  • Basic understanding of cloud applications.
  • No previous ADK experience is required.

  Environment Setup

  Participants will use a Google Cloud project, Vertex AI, BigQuery, Cloud Run, Google ADK, Google Cloud CLI, and a dedicated GitHub repository.

Module 1: Gemini, Prompting, and First Python Application

• Generative AI and Gemini fundamentals
• Vertex AI Studio
• Prompt structure and system instructions
• Structured responses
• Calling Gemini from Python

Hands-on: compare Gemini responses, control output format, and create a small Python application.

Module 2: Building the First ADK Agent

• Agent objective, instructions, tools, sessions, state, and memory
• ADK project structure
• Running and debugging agents locally
• Using the ADK web interface

Hands-on: build, configure, and test a first conversational agent.

Module 3: Tools and Structured Output

• Python functions as agent tools
• Tool descriptions and typed parameters
• Single-tool and multi-tool agents
• Error handling
• Structured output

The repository includes progressive examples for a first agent, calculator, multi-tool agent, and structured calculator.

Hands-on: add calculator and utility tools, test tool selection, and return validated structured results.

Module 4: MCP and BigQuery Analytics Agent

• Introduction to the Model Context Protocol
• MCP servers and agent tools
• Connecting ADK agents to BigQuery
• Asking analytical questions in natural language
• Reviewing generated queries and results

The repository includes a BigQuery analytics agent using a GCP MCP server.

Hands-on: connect an agent to BigQuery through MCP and answer a set of analytical questions.

Module 5: Planner–Executor–Reporter with Dynamic Routing

• Separating planning, execution, and reporting
• Deterministic workflow versus model-driven routing
• Dynamic selection of execution tools
• Using a RAG function as one execution path
• Combining tool results into a final report

The repository includes a planner–executor–reporter workflow where the executor dynamically selects between available tools.

Hands-on: run the workflow, inspect the generated plan, route tasks between a RAG function and another tool, and review the final report.

Module 6: Deployment to Cloud Run

• Preparing an ADK service for deployment
• Dependencies and environment variables
• Deploying from source
• Testing the service URL
• Reviewing logs

The repository includes a simple ADK service prepared for Cloud Run deployment.

Hands-on: deploy the agent application to Cloud Run and test the deployed service.

Capstone Exercise

Participants extend one of the agents created during the day by:

• adding a new tool,
• defining a structured response,
• connecting either BigQuery or the RAG function,
• testing routing behaviour,
• and deploying the application to Cloud Run.