Saoussen Chaabnia

Designing and Deploying Distributed Multimodal Multi-Agent Systems with Google's AI Stac

An AI orchestrator directs a team of specialist agents—a copywriter, designer, and more—to create a full marketing campaign. See the microservice architecture for AI in action.

Designing and Deploying Distributed Multimodal Multi-Agent Systems with Google's AI Stac
#1about 3 minutes

Introducing the AI Creative Studio reference application

A multi-agent system is introduced that turns a campaign brief into a full Instagram campaign using five specialist agents coordinated by an orchestrator.

#2about 6 minutes

Building agents with the Agent Development Kit (ADK)

The Agent Development Kit (ADK) provides a framework for building agents by defining their model, tools, and system instructions.

#3about 4 minutes

Managing domain knowledge with ADK Skills

ADK Skills solve the problem of oversized system instructions by loading domain-specific knowledge on demand, reducing token costs.

#4about 3 minutes

Creating multimodal and orchestrator agents

Agents can be defined by wrapping external tools for multimodal tasks or by using the agent-as-a-tool pattern to create an orchestrator.

#5about 7 minutes

Enabling distributed communication with the A2A protocol

The Agent-to-Agent (A2A) protocol is an open standard that allows agents to communicate and discover each other dynamically across different frameworks and infrastructures.

#6about 3 minutes

Implementing A2A for microservice-style agents

ADK simplifies A2A implementation by converting agents into network-callable servers and creating remote clients with just a few lines of code.

#7about 5 minutes

Connecting agents to external tools with MCP

The Machine-to-Computer Protocol (MCP) standardizes how agents connect to external tools like Notion by using a client-server architecture.

#8about 7 minutes

Deploying agents with Gemini Enterprise Agent Runtime

Gemini Enterprise Agent Runtime is a fully managed, serverless platform for deploying stateful orchestrator agents with built-in scaling, monitoring, and session management.

#9about 11 minutes

Live demo of the multi-agent system in action

A live demonstration shows the deployed system in the Google Cloud console, from submitting a campaign brief to the orchestrator planning its execution.

#10about 4 minutes

Reviewing the final campaign output and resources

The demo concludes by showcasing the final generated campaign assets, including market research, social media posts, images, and a structured Notion project plan.

Related jobs
Jobs that call for the skills explored in this talk.

Featured Partners

Related Articles

View all articles

From learning to earning

Jobs that call for the skills explored in this talk.