DevFest Montréal 2026

← Back to the programme
4-4 · Under PressureEnglish

Building Real-Time RAG Systems with Gemini & the Multimodal Live API

Category
AI & ML
Format
Session — 40 minutes
Audience level
Intermediate

This session will demonstrate how to build a real-time Retrieval Augmented Generation (RAG) system using Gemini 3.7 and the new Multimodal Live API. Participants will explore how to generate both text and audio responses grounded in enterprise documents, with a retail customer support assistant as a working example. We will walk through: - Using Gemini 3.7 Flash, Google DeepMind’s fastest multimodal model, to deliver high-quality, low-latency outputs. - Leveraging the Multimodal Live API for interactive text and audio streaming. - Applying RAG techniques to build grounded, context-aware assistants. - Integrating the solution with enterprise data to enhance customer experiences. By the end of the session, attendees will gain practical experience in developing real-time, multimodal AI applications that improve customer engagement and support; unlocking the power of Gemini 3.7 for enterprise-ready solutions.

Speakers

Olayinka Peter

I am Olayinka Peter Oluwafemi, a TensorFlow Certified ML Engineer and a Google Developer Expert for Machine Learning. With over 7 years of engineering experience, I’ve designed and operated data architectures, built production ML pipelines, and deployed scalable machine learning systems. I enjoy building reliable, production-grade ML systems that solve real business problems. See more of me at olayinkapeter.com