Google Developer Expert
Developer Evangelist
International Speaker
Spoken 213 times in 44 countries
Founder NextAI London
Founder Axiom Masterclass
Stochastic parrot or AGI?
What is a distribution?
Algorithmic Bias
Pick a number between 1 and 10 without saying it out loud.
Rise your hand if you picked the number
7
10
Seven is up to 10 times higher!
x10
7
10
7
10
The Ghost of Out-Of-Distribution
10:10
9:15
Final Boss: Context Contamination
Messi
Ronaldo
Lisbon
Distribution Drift
Original Prompt
New Prompt
Gemini 3
Gemini Evolution to Gemini 3
1h
video
11h
audio
30K
LOC
800
pages
1 Million tokens
NotebookML
Google AI Learning Path
Vertex AI
Complexity
Features
1
2
Generative AI for Developers
Vertex AI
Complexity
Features
Gemini Live in Action
Talk
30 Natural Voices
+24 Languages
2-15min
Show
Video
Attachments
Ask
Code Execution
Function Calling
Web Search
Generative AI for Developers
Vertex AI
Complexity
Features
The AI Playground for Devs
Vibe Code: Building with AI
Chat with Models: Prompting
Experiment with Gemini Live
What is an AI Agent?
Access to Tools and APIs
Python Sandbox
Google Search
Function
Calling
Multi-Agent Systems
Context Agent
Priorisation
Agent
Execution Agent
User
Goal/Rules
Task Queue
Task
Task creation Agent
1. Provide objective
4. Complete task
6. Update tasks
2. Add new tasks
3. Query context
Memory
5. Store task/result
Tools
Gemini Deep Research
AI Voice
Agents
AI Voice Agents Use Cases
Real Time
Native Audio
200-600 ms
30 Voices
Interactive
Half-Cascade
500-800 ms
8 Voices
On Demand
Text-To-Speech
few seconds
30 Voices
Tools: Google Search
Google Search
Run query in
Google Search
Tools: Function Calling
Function Calling vs MCP
Model 2
Model 1
Model 2
Model 1
Demo: AI Voice Agent using MCP
Voice-First Use Cases
24/7 Bookings using AI Voice Agents
Robot Cafe
Useful Links and Materials
Scan to access Slides.
Talk: Building the Next Generation of Voice-First Agentic AI using Gemini 3 Pro
By Gerard Sans
Talk: Building the Next Generation of Voice-First Agentic AI using Gemini 3 Pro
Generative AI has shifted gears. In this talk we explore how Gemini 3 Pro now supports multimodal generation (text, image, audio, video) and real-time agents. You’ll see how Google AI Studio connects the latest model families into a unified workflow of “vibe coding” rather than writing boilerplate. We’ll also dive into the voice-first capabilities made possible via the Gemini Live API: real-time, bidirectional voice and video conversations, tone-aware responses, tool-integration and session memory. Together we’ll look at prompt-to-code flows, media generation, voice-first use cases and the next generation of MCP-driven agentic AI.
