FitSpace

System architecture overview
Backend microservices and data flow
AI agents discovery & execution flow
User dashboard with asset upload
Virtual try-on experience (AI-powered)

Inspiration

When shopping online, there is always doubt: will the clothes really fit me? Will furniture or decor look good in my apartment? I wanted to create a service where AI agents can help remove these uncertainties before purchase.

What it does

FitSpace allows users to:

Upload their own photos (portrait, full body, or interior).
Receive intelligent AI suggestions on how clothes or items can be combined with their photos.
Select a suggestion and get a virtual try-on or interior visualization powered by Google Nano-Banana.

How I built it

Implemented user photo upload and storage.
Designed multi-agent flows using the Agent Development Kit (ADK).
Integrated agent-driven AI suggestions directly on product pages.
Connected a generative model for virtual try-on and product placement in interiors.
Deployed the system on Google Kubernetes Engine (GKE) for scalability and reliability.

Challenges I ran into

Making AI agent suggestions relevant and useful rather than random.
Coordinating multi-step interactions between agents and ensuring smooth data exchange between them.
Building a simple, intuitive interface for photo uploads and AI generations.

Accomplishments that I'm proud of

Built the full cycle solo: from photo upload to agent-powered visualization.
Integrated interactive AI agents that make shopping more engaging.
Deployed the project on GKE, gaining hands-on experience with cloud-native infrastructure.
Delivered a working prototype that shows what the future of e-commerce could look like.

What I learned

User experience starts with small details: even the photo upload flow must be smooth and intuitive.
AI agents can not only enhance visualization but also guide and inspire users with contextual suggestions.
Learned how to containerize and deploy services on Google Kubernetes Engine (GKE), which greatly simplified scaling and orchestration.
Gained first-hand experience with the ADK framework to design and orchestrate AI agents.