Europe 2026
Build & deploy AI-powered apps — Paige Bailey, Google DeepMind
Overview
This talk explores building and deploying AI-powered applications, focusing on the capabilities and accessibility of Google DeepMind's latest models and tools. It highlights the multimodal nature of Gemini models, their integration into AI Studio for rapid prototyping, and the potential for creating complex applications with features like code execution, search grounding, and dynamic world generation. The presentation emphasizes practical demonstrations over slides, showcasing how developers can leverage these advancements for diverse projects.
Who should watch
- AI engineers and developers looking to build and deploy AI applications.
- Product Managers and builders interested in leveraging generative AI for new features and products.
- Anyone curious about the latest multimodal AI models and their practical applications.
- Individuals seeking to understand how to integrate AI tools like Gemini into their development workflows.
Key takeaways
- Google DeepMind has released a suite of powerful AI models, including Gemini 3.1 Pro and Flash, Nano Banana 2 for image generation, an advanced embeddings model, Lyra 3D for music, Genie 3 for world building, and Gemma 4 for open-source applications.
- Gemini models are highly multimodal, supporting various input and output formats like video, images, audio, text, and code, enabling richer application development.
- AI Studio provides a user-friendly interface for experimenting with these models, offering features such as grounding with Google Search, code execution in a sandboxed environment, and URL context for incorporating external information.
- Tools like Gemini Live enable real-time, conversational interaction with AI, supporting screen sharing, video feeds, and dynamic responses in multiple languages, with applications in robotics and augmented reality.
- Project Genie 3 allows for the dynamic generation and navigation of 3D worlds based on descriptive prompts, showcasing a new frontier in AI-driven content creation.
- The Gemma family of open models offers powerful capabilities that can be used via API for free or downloaded for self-hosted fine-tuning and deployment.
- Developers can build and deploy full-stack applications directly within AI Studio, integrating features like databases and authentication, simplifying the path from prototype to production.
Notable quotes
*Gemini models are highly multimodal, supporting various input and output formats like video, images, audio, text, and code.*
*AI Studio provides a user-friendly interface for experimenting with these models, offering features such as grounding with Google Search, code execution in a sandboxed environment, and URL context for incorporating external information.*
Unofficial community note. Prefer the recording for nuance.