A 24-hour Google Cloud Virtual Hackathon centered on Agentic AI invited participants to build applications that think, act, and assist using Gemini, Vertex AI, Firebase, and Google ADK. The winning entry implemented a multi-agent AI system for medical diagnosis and patient care, moving from text-based diagnosis to specialist multi-agent responses and adding multimodal inputs like images. The development process emphasized breaking the problem into actionable technical requirements, prioritizing core features, and designing intelligent agent interactions for real-world scenarios. Advice encouraged commitment, clear execution, storytelling about the idea, and building a foundation with product potential beyond the event.
A comprehensive multi-agent AI system for medical diagnosis and patient care, built with Google ADK and specialized medical agents. Thoughts on doing the Hackathon The Agentic AI theme made it very different from the web development hackathons I'd done in the past. It wasn't just about building fast - it was about designing intelligent behavior and thinking through how autonomous agents could interact in real-world scenarios.
My approach started with breaking down the problem statement into actionable technical requirements...I prioritized core features such as text-based diagnosis and then expanded the system by scaling from a single-agent model to a multi-agent setup with specialist responses. I also began incorporating support for images and multimodal inputs, aiming to make the assistant more comprehensive over time...My mindset from the beginning was to build not just for the hackathon, but to create a strong foundation for a product that could grow well beyond it.
Collection
[
|
...
]