Beyond 5G: Engineering Low-Latency USSD & Voice Workflows with FastAPI & Cloud Run

When designing backend systems, modern cloud engineers frequently take high-bandwidth, low-latency client environments for granted. However, millions of end users globally operate under severe digital, economic, and infrastructural constraints. To reach these populations, developers must bypass heavy client-side applications and interface directly with core telecommunications primitives: USSD (Unstructured Supplementary Service Data) and IVR (Interactive Voice Response).

Using a real-world public health surveillance platform as an engineering case study, this workshop provides a step-by-step breakdown of how to build, containerize, and deploy an enterprise-grade cellular ingestion engine on Google Cloud Platform.

Attendees will learn how to solve the unique distributed systems challenges imposed by telecommunications gateways—such as rigid 3-second backend timeouts, transient session states, and duplicate webhook retries caused by volatile cellular towers.

Key Technical Takeaways:

Stateless Microservice Orchestration
Deterministic Session Tokenization
Idempotent NoSQL Ingestion
Edge Media Delivery for Voice (IVR)
Asynchronous Event Loops

Build & Deploy AI-Powered Web Application on Google Cloud Run Using AI Studio

In this workshop, you will learn how to build and deploy a web application using the "vibe coding" capabilities of Gemini in Google AI Studio. You will create "Snake & Beats"—a retro Snake game with an integrated music player and neon aesthetic—starting from a single natural language prompt. We won't just talk about the future of development; we will demonstrate it by having every person in the room deploy a live application to the cloud.

The lab guides you through the following key steps:

Rapid Prototyping: Generate a functional React application using Gemini 3.0 Pro in AI Studio's Build Mode.
Multimodal Refinement: Iterate on your design using voice commands and Annotation Mode, which allows you to draw directly on the UI preview to communicate visual changes.
Persistent Context: Use System Instructions to maintain a consistent "coding vibe" and visual persona (e.g., Glitch Art style) throughout your session.
Deployment & CI/CD: Bootstrap a GitHub repository directly from AI Studio and deploy your live application to Google Cloud Run with a single click.
This session is a celebration of diverse ideas, proving that with Gemini and Google Cloud, the only limit to software development is the clarity of your vision.

Build & Install a Native Android App Using AI Studio

With the monumental updates announced at Google I/O 2026, the barrier to mobile development has completely vanished. Google AI Studio’s completely overhauled Build Mode has expanded beyond simple web mockups to natively support full-stack Android application development directly in the browser via natural language prompting.

In this fast-paced, highly visual live-coding session, we will skip the heavy local SDK setups, Gradle configurations, and IDE downloads. Instead, we will leverage Gemini 3.5 Flash inside AI Studio to prompt, iterate, and deploy a native Android utility from scratch.

Attendees will see the AI generate production-ready Kotlin and Jetpack Compose (Material 3) code, use the new Annotation Mode to sketch design changes directly onto an embedded browser-based Android emulator, and see how WebUSB allows instant on-device installation. Finally, we’ll demonstrate how advanced developers can seamlessly transition these cloud-built states directly into Android Studio using its new Agent Skills.

Objectives:

Mastering AI Studio Build Mode: Understand how to structure natural language prompts to guide Gemini 3.5 Flash in generating robust, idiomatic Kotlin and Material 3 layouts.
Rapid Multi-modal Prototyping: Learn to use the new native in-browser Android Emulator and Annotation Mode to test, debug, and visually iterate on apps without local environments.
Bridge to Production: Discover the seamless handoff workflow from cloud-based "vibe coding" to Android Studio.

Build and Deploy Intelligent No-Code Agents with Gemini Enterprise

Learn how to configure, build, and deploy intelligent, no-code AI agents using the Gemini Enterprise platform. In this hands-on workshop, you will set up a Google Cloud project and configure a dedicated Gemini Enterprise tenant application. You will configure identity authentication using Google Identity and customize configurations to activate the Agent Designer. Next, you will write natural language prompts to create and dynamically modify a functional AI agent inside the Agent Designer's interactive Flow view. Finally, you will test your agent's conversational capabilities in real-time and share it with your organization via the integrated Agent Registry

Build and Deploy Your Portfolio Website with Google AI Studio and Cloud Run

Where does building with AI start today? It begins with a simple query or prototype. In this hands-on workshop, you will learn how to use Google AI Studio's Build Mode (often called "vibe coding") to design, test, and prototype a professional, responsive personal portfolio website using natural language driven by Gemini models. After polishing your frontend layout in AI Studio's interactive preview sandbox, you will deploy your application as a serverless container directly to Cloud Run with a single click.

What you will learn:

Prompting & Prototyping: How to prompt Google AI Studio's Build mode to generate, style, and iterate on a modern frontend layout using natural language.
Testing & Troubleshooting: How to run validation checks and automatically diagnose or repair errors using the interactive preview and AI-assisted "Fix" features.
Serverless Deployment: How to deploy your live web application directly to Cloud Run from the AI Studio user interface.
Custom Domains (Optional): How to configure DNS records and map a custom domain to secure your live site

Build a Gemini-Powered YouTube Summarizer

YouTube contains a vast amount of educational and informational content, but watching long videos can be time-consuming. A YouTube summarizer can help users quickly extract the key ideas from a video without needing to watch the entire recording.

In this hands-on workshop, you will build a web application that generates concise summaries of YouTube videos using Google Gemini. The application will accept a YouTube video URL, process the video's transcript, send the relevant content to the Gemini API, and display an AI-generated summary to the user.

By completing this workshop, you will gain practical experience integrating generative AI into a web application. You will learn how to work with the Gemini API, use the Google Gen AI SDK, handle user input and video content, and present AI-generated results through a simple web interface.

Learning Outcomes

By the end of this workshop, you will be able to:

Build a functional web application that summarizes YouTube videos.
Integrate the Gemini API into an application.
Use the Google Gen AI SDK to interact with Gemini models.
Process video transcripts as input for an AI model.
Generate concise, useful summaries using Gemini.
Connect an AI-powered backend with a web-based user interface.
Handle API responses and display generated content to users.
By the end of the workshop, you will have a working Gemini-powered application that demonstrates how generative AI can be integrated into a practical real-world use case

Build & Deploy Serverless AI Pipelines on Google Cloud with Antigravity

Take your AI agents to the cloud! You will use Google Antigravity to autonomously architect, build, and deploy a serverless event-driven document processing pipeline. Together with the AI, you will build a system that ingests files from Cloud Storage, processes them with Cloud Run and Gemini, and stores metadata in BigQuery.

Key Takeaways:

1. AI-Assisted Architecture: Use Antigravity's Planning mode with Gemini models to outline complex Google Cloud architectures and task lists.
2. Infrastructure Generation: Instruct the agent to generate shell scripts that provision Cloud Storage buckets, Pub/Sub topics, and BigQuery datasets.
3. Microservice Deployment: Oversee the AI as it writes and deploys a Python-based Cloud Run application integrated with Gemini on Vertex AI.
4. Automated Verification: Use Antigravity’s Walkthrough artifacts to test and validate the end-to-end data pipeline automatically.

Building and Deploying a Multi-Agent "Zoo Tour Guide" on Cloud Run with Google ADK

Dive into the world of intelligent, tool-using AI agents by building a multi-agent system from scratch! In this hands-on session, we will focus on the code and architecture required to build a "Zoo Tour Guide" agent using the Google Agent Development Kit (ADK). Instead of relying on a single generic AI, attendees will learn how to build a specialized team of agents, including a "Researcher" to find facts and a "Formatter" to polish the final answer. We will explore how to give these agents access to the outside world by connecting them to external tools like the Wikipedia API, and how to orchestrate their interactions for a seamless user experience.

Key Learning Points:

1. Structuring for the ADK: Learn how to properly structure a Python project and set up a development environment using uv for ADK deployment.
2. Implementing Tool-Using Agents: Discover how to equip your AI with external capabilities, such as integrating the LangchainTool to allow your agent to query Wikipedia for general world knowledge.
3. Managing Agent Memory: Understand how to use ToolContext to capture and save user prompts into the agent's short-term memory (state), allowing data to be shared seamlessly between different agents in the workflow.
4. Designing Multi-Agent Workflows: Master the use of SequentialAgent to act as a "back-office manager," automatically passing shared memory and executing specialized sub-agents in a reliable, fixed sequence.

Making Your Google ADK Agents Talk to the World

You’ve built a brilliant AI agent, but it can’t just live in a local testing browser forever. In this session, we’ll explore how to make your Google Agent Development Kit (ADK) creations truly connect with external users and systems by moving beyond the default adk web interface. Attendees will learn how to spin up a local REST API using the adk api_server command, enabling web applications, mobile app backends, and microservices to communicate with their agents over standard HTTP requests. We will also dive into programmatic execution in Python, demonstrating how to natively embed agents into custom applications or data pipelines using the Runner class and InMemorySessionService. Finally, we'll look at using adk run for terminal-based execution, perfect for server environments without a GUI or automated CI/CD workflows. Whether you are building an independent API service or seamlessly embedding AI into an existing app architecture, this workshop will give you the deployment skills to take your agents from isolated prototypes to real-world solutions.

Building and Customizing Your First AI Agent with Google's ADK

Step into the world of AI agents with the Google Agent Development Kit (ADK). In this foundational workshop, developers will learn how to set up their local Python environment, install the ADK, and securely configure API keys using either Google AI Studio or Vertex AI. We will dive deep into the four core parameters that define every agent's identity: model (the reasoning engine), name (the unique identifier), description (for multi-agent routing), and instruction (the behavioral blueprint). Through hands-on exercises, participants will learn how to transform a generic default assistant into a highly specialized persona, such as a patient algebra tutor, and instantly test its behavior using the local adk web visual interface. You will leave this session with a fully functional agent and the foundational skills to guide LLM behavior.

AI-Driven Design-to-Code with Google Stitch and Antigravity

In this hands-on workshop, we will bridge the gap between high-fidelity AI-driven design and agent-first development to build a production-ready website. Join me as we explore how to seamlessly translate visual design metadata into modular code using modern autonomous agents.

During this session, you will learn how to rapidly prototype a full-scale web design using natural language and the Gemini 3 model in Google Stitch. We will then move into the Antigravity IDE, configuring a secure bridge using the Model Context Protocol (MCP) to allow our autonomous agent to fetch your project's "Design DNA" directly from Stitch. Finally, you will guide the agent to autonomously scaffold and build a pixel-perfect React and Tailwind CSS application, using the IDE's integrated browser to "Vibe Check" and instantly refine the code against your original design.

Key Takeaways:

How to use Google Stitch to generate high-fidelity UI prototypes using natural language.
How to configure MCP Servers within the Antigravity IDE, giving the AI secure, real-time context from external tools.
How to command autonomous agents to extract design tokens (like color palettes and typography) and translate them into a modular React codebase.

Prerequisites for Attendees: To follow along, please come prepared with a Chrome browser, Node.js (v18+) installed locally, an active Google Cloud Project with billing enabled, a Google Stitch account, and the Antigravity IDE installed.