Google DeepMind · Developer Experience

Omar Sanseviero

DevX Lead, Google DeepMind

From Hugging Face Chief Llama Officer to leading Gemma, AI Studio & the Gemini API developer experience. Building the future of open AI.

Gemma Lead150M+ Model DownloadsO'Reilly AuthorEx-Hugging Face CLO

150M+

Gemma Downloads

Token Context (Gemini 3)

140+

Languages Supported

27B

Largest Gemma Model

Shifting to an Era of Action

The three generations of Gemini, each representing a fundamental shift in capability.

Gemini 1

Understanding

Processing and comprehending multimodal inputs

Gemini 2

Thinking

Deeper reasoning and chain-of-thought capabilities

Gemini 3

Action

Autonomous execution, function calling, multi-step workflows

Google DeepMind Full Stack

Five layers of infrastructure powering the next generation of AI applications.

Foundation Models

Gemini, Gemma, Imagen, Lyria, Veo

AI Frameworks

JAX, Keras

Developer Tools

Google AI Studio, Colab, Kaggle

Infrastructure

Vertex AI, TPUs, GPUs, CPUs

Deployment

Edge AI, Web, Mobile, Browser

Beyond Gemini Models

A comprehensive portfolio of AI products from Google DeepMind.

Image Generation

Imagen4 / Nano-Banana 2

Video Generation

Veo 3

Music Generation

Lyria

Object Detection

Gemini Robotics

Autonomous Agents

Mariner & Jules

Open Models (150M+ downloads)

Gemma 3

Scientific Discovery

AlphaEvolve

Ask About the Presentation

Chat with Gemini + generate images with Nano-Banana 2. Ask questions or try the suggestion bubbles.

Gemini Chat + Nano-Banana 2

Ask questions or generate images

Hi! I'm powered by Gemini and know everything about Omar's presentation at Google for Startups Campus Warsaw. I can also generate images using Nano-Banana 2! Try the suggestions below or ask anything.

Talk to Omar's AI Voice

Get a phone call from an AI voice agent that knows everything about Google DeepMind's developer experience.

Call Omar's AI

Enter your phone number and receive a call from Omar's AI assistant powered by ElevenLabs + Twilio

Phone Number (with country code)

Powered by ElevenLabs Conversational AI + Twilio. The AI voice agent knows about Google DeepMind, Gemma, Gemini API, and Omar's presentation topics.

Frequently Asked Questions

Key takeaways from Omar's presentation at Google for Startups Campus.

What is Gemma and how does it differ from Gemini?

Gemma is Google's open-weight model family available for download and local deployment. Gemini is Google's proprietary frontier model accessed via API. Gemma 3 supports 5 sizes (270M to 27B), 128K context, multimodal input, and 140+ languages.

What is the Interactions API shown in the demo?

The Interactions API is the new agent-building API for Gemini, demonstrated in Omar's agents_warsaw.ipynb Colab notebook. It enables multi-turn agent conversations with function calling and tool use.

What is Search Grounding?

Search Grounding allows Gemini to ground its responses in real-time Google Search data. Omar demonstrated this with an isometric 3D illustration of Warsaw's Old Town that included live weather data (11°C, cloudy).

What is AI Studio Build?

AI Studio Build (ai.studio/build) is the new mode in Google AI Studio where developers can create full applications directly. Omar demoed 'Llama Drama', a whimsical app built with Gemini 3 Flash Preview.

How can I get Google Cloud startup credits?

Google for Startups Cloud Program offers up to $350K in credits: $2K Bootstrap tier, $25K Accelerator, $250K AI validation for Pre-Seed/Seed, and Committed Use Discounts for Series A+.

What is Gemma 3n and its use case?

Gemma 3n is the mobile-first variant using MatFormer and Per-Layer Embeddings (PLE) for low RAM usage. It supports 32K context, audio+text+visual input, and runs ~1.5x faster than predecessors. Ideal for privacy-first, offline-ready on-device AI.

Build with Google DeepMind

Start exploring Gemma, the Gemini API, and AI Studio today. Open models, powerful APIs, and the tools to build the next generation of AI applications.

Open AI Studio Explore Gemma