Nuvona

Nuvona empowers and educates first-time homeowners to plan ahead by predicting insurance costs, climate risks, and property trends using AI, transforming raw data into an easily ascessible interface.

Comment

Inspiration

Over 1 million Florida residents currently live without home insurance, and rising non-renewal rates are leaving more families unprotected and uninformed. As a team of immigrants and first-generation Americans, we've seen firsthand how homeownership can be harder to navigate for families that are new to the U.S. So we wanted to create a tool that makes that journey to financial literacy easier for others. That is where the idea of Nuvona was born, a mix of the words new/nouveau and vona, which in Icelandic means hope (A New Hope). We hope to share an experience that builds a bridge between the complex system of housing insurance and the people who need clarity most.

What it does

Nuvona not only predicts insurance costs, property trends, and climate susceptibility in chosen areas using AI, but also helps first-time homeowners understand these prices and trends and prepare for future risks through concise, visual insights.

How we built it

We designed Nuvona as a React and TypeScript front-end run by a Python Flask back-end.

  • Frontend: React, Tailwind, and Figma for UI components, integrated with an interactive map built using OpenStreetMap API, and Vite, Leaflet, and Lucide-React for build tools.
  • Backend: FastAPI + Uvicorn for fast server handling, Pydantic for data validation, Google GenAI + Google Cloud AI Platform for AI predictions, osmnx + geopandas + shapely for mapping, pandas + pyarrow for data processing/transfer, and requests for data fetching and API communication.

Challenges we ran into

  • Figuring out how to use Gemini Keys
  • Learning how to integrate an OpenStreetMap API
  • 2/4 of us were first-time hackers, so learning everything about the backend and frontend of our website and how to navigate around using the terminal (like cd .., npr run dev, adk run, etc.) was a challenge.
  • Implementing AI into our ChatBot with prompt-engineered responses to guide users through house insurance.

Accomplishments that we're proud of

  • Using Google ADK for the first time and actually be able to implement our AI ChatBot from the backend to our frontend.
  • First time working with an API (OpenStreetMap) and utilizing its map features.
  • Having a fully functional end-to-end app.

What we learned

  • How to design frontend using Figma.
  • How to manage databases with backend environment.
  • How to train A2A communication/interaction.
  • How to connect backend and rontend to create full stack application.
  • How to engineer prompts for AI model using google-adk.
  • Redbull has an interesting flavor.

What's next for Nuvona

We plan to expand this project and explore other types of insurances such as commercial insurance, and even expand the whole map nation-wide.

Built With

Share this project:

Updates