| Intro & start with Anna Muzykina | Workshop to activate cloud credits and prepare your environment to workshops. |
| Keynote speaker Saheed Adewumi: From Prompt to Purpose: Navigating the Dawn of the Agentic Era | For the last two years, we have lived in the era of Generative AI—a world of chatbots and creative assistants. But a more profound shift is underway.
We are moving from AI that suggests to Agents that execute. In this 15-minute keynote, Saheed Adewumi explores the "Agentic Era," where AI systems possess the autonomy to reason, use tools, and achieve complex goals with minimal human intervention. |
| Speaker Lesly Zerna: Gemini, AI Studio, Agent Platform (formerly VertexAI) - start building today! | Let's explore Google Tools to get started in the GenAI & Agentic Era.
I'll showcase demos and real life examples on how to build agents using Gemini as the brain, tools and also low code Agent Platform. |
| Speaker Geeta Kakrani: Cost-Efficient Multi-Agent Systems with Dynamic Routing & Caching | Modern AI applications require multiple agents, models, and intelligent orchestration—but costs and latency can quickly spiral out of control.
In this workshop, we’ll build a multi-agent AI system that uses dynamic routing, semantic caching, and vector search to optimize performance and cost. You’ll learn how to decide when to call large vs small models, reuse responses using caching, and leverage vector databases for context-aware retrieval.
By the end, you’ll have a clear architecture for building scalable, cost-efficient Agentic AI systems using modern tools and frameworks. |
| Speaker Aditya Joshi: Building AI Agents with Gemini & LangChain in TypeScript | LLMs become truly powerful when they can reason, use tools, and take actions - not just respond. In this session, we build a fully functional AI agent using Google's Gemini models and LangChain's TypeScript SDK, from scratch.
You'll learn how LangChain's agent abstraction works under the hood — how it binds tools to a model, parses reasoning steps, and loops until a goal is reached. We'll use Gemini's native function-calling to drive tool selection, compose a real-world agent with multiple tools, and cover patterns for memory, streaming, and making your agent reliable in production.
TypeScript developers are increasingly reaching for LLMs, but most stop at simple chat completions. This talk shows the next step: building a goal-driven agent that thinks in steps, calls tools, and handles real tasks — using LangChain.js and Gemini, two of the most widely adopted tools in the space. Attendees leave with working TypeScript code and a clear mental model for building agents they can actually ship. |
| Speaker Gabriel Preda: Build healthcare assistant with VertexAI, MedGemma, and Gemini | Discover how to build an healthcare assistant with VertexAI, MedGemma, Gemini.
Use Antigravity to get from a simple medical image interpreter to a healthcare assistant implemented as an agentic application powered by multiple multi-modal models, including guardrails, orchestration and validation. |
| Speaker Chukwuemeka Chukwurah: Chatbots to AI Teams: Building Distributed Multi-Agent Systems on Google Cloud | Build more than a chatbot—build a team of AI specialists. In this hands-on codelab, participants will create a distributed multi-agent Course Creation System using Google’s Agent Development Kit (ADK).
Instead of relying on a single LLM, attendees will design specialized agents for research, evaluation, content generation, and orchestration, each with a clear role in solving complex tasks collaboratively.
Along the way, they’ll learn tool use with web search, structured outputs with Pydantic, Agent-to-Agent (A2A) communication, iterative feedback loops with LoopAgent, and workflow orchestration with SequentialAgent.
The session concludes with deploying the full system to Google Cloud Run—showing how scalable, production-ready AI agents can move beyond demos into real-world applications. |
| Speaker Nicolas Molina Monroy: Generative UI: Interfaces that create themselves (Angular + Hashbrown) | What if your application stopped being static and started adapting to the user in real-time?
In this session, we will explore how Gemini AI can be integrated into Angular + Hashbrown to build dynamic, intelligent, and contextual interfaces. We'll dive into the concept of Generative UI, demonstrating that the front end is no longer just for rendering—it now thinks and evolves. |
| Speaker Amira Best: AI Agent Evaluations | As AI agents tackle increasingly complex tasks, rigorous evaluation is essential for building reliable systems.
This presentation will cover the core types of evals, the trade-offs between offline and online approaches, and how to build high-quality golden datasets with well-aligned LLM judges.
We'll close with how a production flywheel turns real-world signals into continuously improving evaluation pipelines. |
| Speaker Leong Lai Fong: Deploy, Manage and Observe ADK Agent on Cloud Run | Agent Development Kit (ADK) empowers you to create agents capable of complex, multi-agent workflows. By leveraging Cloud Run, a fully managed serverless platform, you can deploy your agent as a scalable, containerized application without worrying about the underlying infrastructure. This powerful combination allows you to focus on your agent's core logic while benefiting from Google Cloud's robust and scalable environment.
In this session, you will explore the seamless integration of the ADK with Cloud Run. You will learn how to effectively deploy, manage, and monitor your ADK-powered agents on Cloud Run. Comprehensive guide includes:
- Initial deploy the agent to Cloud Run
- Configure and Test Cloud Run auto scaling with load testing
- Strategy to gradual release with Cloud Run
- Setup ADK Agent tracing to Cloud Trace and inspect agent run tracing |
