Building GPU-accelerated multi-agent apps with Google ADK and Gemma 4

Published on 12 May 2026, 17:00

Join the Google Cloud & NVIDIA community → goo.gle/google-nvidia-programs

On this week’s livestream, join Chelsie Czop and @NVIDIA's Jay Rodge and learn how to build a GPU-accelerated sustainability intelligence app from scratch.

You’ll learn how to orchestrate a team of specialist agents using the open-source Google Agent Development Kit (ADK), serve Gemma 4 efficiently on Cloud Run using NVIDIA RTX PRO 6000 GPUs, and connect them with Milvus for policy retrieval.
Whether you are looking to deploy local models or scale production agentic workflows, this is the architecture blueprint you need.

This livestream originally aired on May 12, 2026 at 9:00 A.M. PDT / 12:00 P.M. EDT.

Chapters:
0:00 - Countdown
1:48 - Intro
3:58 - What are AI agents enabling?
13:40 - G4 powered by NVIDIA RTX PRO 6000 Blackwell GPUs
18:28 - GPU accelerated sustainability intelligence
21:32 - [Demo] Sustainability Agent
27:43 - What is Model Context Protocol (MCP)?
30:03 - Where is agentic AI being used for workloads?
36:56 - How can developers stop AI agents from getting stuck in a loop?
39:39 - When should organizations transition from human-in-the-loop to agentic AI?
42:02 - Does Gemini Enterprise have access to GPUs?
42:50 - What challenges do AI agents face when retrieving accurate policy data?
46:13 - How do organizations approach security, privacy, and regulations when multi-agent orchestration is used?
49:08 - Where is Jay Rodge using AI agents?
53:32 - How to get started with agentic AI on GPUs
56:09 - Wrap up

Speakers: Chelsie Czop, Jay Rodge
Products: Agent Development Kit, Gemma 4, Cloud Run