Build with AI: Hands-on workshop with Google

Building Multi-Agent AI Systems with Google ADK was a deep-dive session designed for developers and AI practitioners eager to design and deploy advanced multi-agent systems using Google’s Agent Development Kit (ADK) and related services. The session focused on practical approaches to building collaborative AI agents capable of solving complex, domain-specific problems through intelligent task allocation and coordinated decision-making.

Participants explored how modern agentic architectures enable teams to go beyond single-model workflows and instead orchestrate multiple agents working together. Through guided explanations and live demonstrations, attendees gained clarity on how to structure, deploy, and scale multi-agent systems for real-world use cases.

With a strong hands-on component, the session equipped attendees with actionable insights into architecting agent workflows and applying them to scenarios such as data analysis, risk assessment, and decision aggregation.

Highlights

The session delivered a comprehensive walkthrough of multi-agent concepts and architecture patterns, showcasing how Google ADK can be used to build intelligent, collaborative systems. Attendees learned how different agent types and workflows could be combined to tackle complex problems efficiently and reliably.

Below is a summary of the key topics covered during the session:

Multi-Agent Concepts & Architecture:

• Understanding collaborative AI and coordinated decision-making.
• Designing sequential, parallel, and hierarchical agent workflows.
• Choosing the right patterns for domain-specific problems.
• Structuring systems for scalability and reliability.

Agent Types with Google ADK:

• Working with SequentialAgent for step-by-step reasoning.
• Using ParallelAgent for concurrent task execution.
• Leveraging LLMAgent for intelligent reasoning and generation.
• Orchestrating agents for efficient task allocation.

Hands-On Demonstration:

• Building a multi-agent investment analysis system.
• Applying agents to data analysis and risk assessment.
• Aggregating decisions from multiple agents for better outcomes.
• Exploring real-world applications of agentic AI.

Advanced Patterns:

• Implementing generator–critic loops for iterative improvement.
• Designing human-in-the-loop workflows for control and trust.
• Enabling secure inter-agent communication.
• Managing coordination in complex agent ecosystems.

Beyond the technical learning, the session provided a platform for participants to discuss emerging agentic patterns, exchange ideas, and connect with others exploring advanced AI system design.

Designed for developers and AI builders ready to push the boundaries of intelligent systems, attendees left with practical frameworks and confidence to start building their own multi-agent solutions using Google ADK.

Stay tuned for more sessions as we continue to explore the future of agentic AI and collaborative systems!