In this course, you’ll learn how to estimate and manage GenAI solution costs, focusing on the unique unit economics of AI versus traditional cloud billing. You will learn to: Understand Metering: Master the four Google Cloud GenAI metering categories and the token economy driving costs. Estimate Costs: Build defensible estimates using the Pricing Calculator for models, context caching, and agentic components. Optimize Spend: Leverage controls—such as input/output token reduction, caching, RAG efficiency, and model routing—while balancing cost, quality, and latency. Implement Governance: Apply FinOps best practices, including budgets, quotas, cost attribution, and Provisioned Throughput management to control spend.
In this course, you'll learn the disciplined process of hill climbing to transform your generative and agentic systems into reliable, high-performing tools. You'll examine the iterative 8-step loop by diagnosing system failures through trajectory analysis and applying precise interventions across the model, prompt, tool, and framework layers. By the end of this course, you'll be prepared to optimize system architecture and automate the hill climbing cycle while balancing quality, cost, and latency as first-class metrics.
In this course, you'll learn how to establish a rigorous upgrade regression testing pipeline, safely deploy models using A/B testing and/or shadow mode, and address legacy prompt technical debt. You'll also learn the latest about Gemini 3 parameters including thinking levels and thought signatures, as well as new capabilities like media resolution controls and streaming function calls.
In this course, you learn to evaluate, diagnose, and optimize AI agents on the Gemini Enterprise Agent Platform (GEAP). You begin where most teams begin: the agent runs, but you have no eval cases, no test data, and no production traffic to grade it with. From there you follow the Quality Flywheel, the evaluate-analyze-optimize loop at the center of GEAP. You instrument the agent so it emits the telemetry GEAP reads, generate eval cases by simulation, choose the metrics that grade them, run offline evaluations, monitor live traffic and alert on quality drift, then cluster failures and optimize. A final module covers build-time evaluation with the Agent Development Kit (ADK), which runs on your machine before you deploy.
Throughout this course, you'll learn how to establish rigorous evaluation criteria, perform evaluations, design objective rubrics, calibrate autoraters, and simulate evaluation data. You'll gain the skills needed to design, execute, and scale a comprehensive evaluation plan that aligns system capabilities with organizational KPIs. This course is designed for technical practitioners, machine learning engineers, and software architects who build and deploy generative and agentic applications.
In this course, you explore the Code Execution feature of the Gemini Enterprise Agent Platform, which lets AI agents safely generate and run Python code in isolated sandbox environments. You learn how the Agent Sandbox fits into the broader platform architecture, how to configure and operate Code Execution sandboxes using the Agent Platform SDK, and how to integrate code execution into agent workflows with the Agent Development Kit (ADK).
In this course, you'll learn to use the Agent Development Kit (ADK) to build systems where multiple AI agents collaborate on complex tasks. You'll start with the ADK agent model: how ADK represents agents, tools, and runners, and how a single agent is configured and run. You'll then make tools the model can call, persist session state across agents, and instrument the execution lifecycle with callbacks and plugins. Next, you'll orchestrate multiple agents using ADK's template workflow agents and graph-based workflows, and ground them in enterprise data through multi-source retrieval and MCP integrations. Finally, you'll deploy a multi-agent system to Agent Runtime as a managed service and register and share it through Gemini Enterprise so users across an organization can reach it.
In this course, you'll learn to build and run enterprise agents on the Managed Agents API, the managed agent runtime on Gemini Enterprise Agent Platform. Three conceptual lessons give you the mental model. You'll learn why a real business task needs an agent rather than a chat model. You'll examine how the platform splits into a control plane that defines agents and a data plane that runs them. You'll learn how to assemble an agent from a definition, a sandboxed environment, mounted data, tools, and skills. And you'll learn how to run it with background interactions, a streamed reason-act loop, resilient typed results, and state that persists across turns. You then put that model to work in a hands-on lab, where you build, run, and harden a retail merchandising agent for Cymbal Retail from an empty project to a production-shaped deployment. By the end, you'll be able to design, build, and operate a managed agent of your own.
In this challenge lab, you will act as a cloud engineer supporting the Cymbal Pools finance team. Your mission is to deploy a BigQuery-enabled agent to Agent Runtime to help process invoice data using natural language. Rather than building from scratch, you inherit an unsecured deployment. You must establish basic data governance by configuring the Agent Development Kit (ADK), deploying the agent with a dedicated SPIFFE identity, identifying permission blocks, and applying least-privilege IAM roles so the agent can safely query and update the BigQuery database from the Agent Runtime Playground.
In this challenge lab, you will demonstrate your ability to add agents to a Gemini Enterprise app. You will build an agent with Agent Designer. And you will build a no-code agent with Agent Development Kit, deploy it to Agent Engine, and add it to the Gemini Enterprise app.
In this challenge lab, you act as a Security Engineer deploying a secure Gemini Enterprise environment for Cymbal Bank. You will ground Gemini in web search and internal Workspace sources to ensure accurate, contextual responses. To maintain compliance, you will configure Model Armor policies to filter sensitive data and block threats like prompt injections and malicious URLs. Finally, you will manage specific end-user features to customize the AI experience safely
Complete the Evaluate and Improve Agent Development Kit Agents skill badge to demonstrate your ability to use ADK's evaluation tools to "hill climb" — making measurable, iterative improvements to an agent. You will run an initial evaluation to establish a baseline, apply optimization techniques, and re-evaluate the agent to measure your success.
In this challenge lab, you will demonstrate your ability to author agents using Agent Development Kit (ADK), deploy those agents to Agent Engine, and use them from a web app. Complete the challenge lab to earn a Google Cloud skill badge.
Complete the Accelerate Development with Antigravity skill badge to demonstrate your proficiency in using the Antigravity IDE for developing agentic workflows. You will be tasked with configuring an MCP server, authoring custom agent skills and rules, prototyping with the Agents CLI, and deploying to the Google Cloud Agent Runtime. A skill badge is an exclusive digital badge issued by Google Cloud in recognition of your proficiency with Google Cloud products and services and tests your ability to apply your knowledge in an interactive hands-on environment. Complete the assessment challenge lab, to receive a skill badge that you can share with your network. When you complete this course, you can earn the badge displayed here and claim it on Credly! Boost your cloud career by showing the world the skills you have developed!
In this course, you’ll learn to simplify the creation of autonomous enterprise agents using the Agent Development Kit (ADK) and you will learn how to leverage the power of Antigravity and the Agents CLI to transform your agent development workflow. In this course. You will explore practical techniques to automate repetitive tasks and significantly accelerate the creation of robust, enterprise-grade agents, ensuring scalability and efficiency in your AI projects.
Aprende a usar Gemini Notebook para crear una guía de estudio personalizada para el examen de certificación Professional Cloud Architect. Repasarás las funciones de Gemini Notebook, crearás un cuaderno en Gemini Notebook y aprenderás a usar una guía de estudio como práctica para un examen de certificación.
Agentes de IA generativa: transforma tu organización es el quinto y último curso de la ruta de aprendizaje Líder de IA generativa. En este curso, se analiza cómo las organizaciones pueden usar agentes de IA generativa personalizados para abordar desafíos empresariales específicos. Puedes obtener experiencia práctica a través de la creación de un agente de IA básico mientras exploras los componentes de estos agentes, como los modelos, los bucles de razonamiento y las herramientas.
Apps de IA generativa : transforma tu trabajo es el cuarto curso de la ruta de aprendizaje de Líder de IA generativa. En este curso, se presentan las aplicaciones de IA generativa de Google, como Gemini para Workspace y NotebookLM. Te brinda orientación sobre los conceptos como la fundamentación, la generación mejorada por recuperación, la creación de instrucciones eficaces y el desarrollo de flujos de trabajo automatizados.
IA generativa: explora el panorama es el tercer curso de la ruta de aprendizaje de Gen AI Leader. La IA generativa está cambiando la manera en la que interactuamos y trabajamos con el mundo que nos rodea. Pero, como líder, ¿cómo puedes aprovechar su poder para generar resultados comerciales reales? En este curso, explorarás las diferentes capas del desarrollo de soluciones de IA generativa, las ofertas de Google Cloud y los factores que se deben considerar cuando se selecciona una solución.
IA generativa: descubre los conceptos fundamentales es el segundo curso de la ruta de aprendizaje de Gen AI Leader. En este curso, descubrirás los conceptos fundamentales de la IA generativa explorando las diferencias entre esta, el AA y la IA, y comprendiendo cómo los diferentes tipos de datos permiten abordar desafíos empresariales con la IA generativa. También conocerás las estrategias de Google Cloud para abordar las limitaciones de los modelos de base y los desafíos clave para desarrollar e implementar la IA de forma responsable y segura.
IA generativa: más allá del chatbot es el primer curso de la ruta de aprendizaje de Gen AI Leader y no tiene requisitos previos. El objetivo de este curso es profundizar los conocimientos básicos sobre chatbots para explorar el verdadero potencial de la IA generativa para tu organización. Explorarás conceptos como los modelos de base y la ingeniería de instrucciones, que son fundamentales para aprovechar el poder de la IA generativa. Este curso también te sirve como guía para las consideraciones importantes que debes tener cuando desarrollas una estrategia de IA generativa exitosa para tu organización.
Completa la insignia de habilidad introductoria Prepara datos para las APIs de AA en Google Cloud y demuestra tus habilidades para realizar las siguientes actividades: limpiar datos con Dataprep de Trifacta, ejecutar canalizaciones de datos en Dataflow, crear clústeres y ejecutar trabajos de Apache Spark en Managed Service for Apache Spark y llamar a APIs de AA, como la API de Cloud Natural Language, la API de Google Cloud Speech-to-Text y la API de Video Intelligence.