Inspiration

  • Observing how poorly resources and infrastructure are distributed across Buffalo neighborhoods inspired this project.

What it does

  • An AI-powered analysis tool for Buffalo’s neighborhoods that visualizes infrastructure, socioeconomic trends, and poverty-related insights at the ZIP code level.
  • Displays ZIP-level maps and charts showing trends over time and comparisons.
  • Predicts poverty rates for hypothetical scenarios through a simulation interface.
  • Incorporates Gemini to answer user questions

How we built it

  • Machine learning with PyTorch and Scikit-Learn.
  • Backend using FastAPI and MongoDB to serve ZIP-level and historical data.
  • Frontend with React and Leaflet for interactive maps and charts.

Challenges we ran into

  • Collecting consistent and reliable data across all ZIP codes.
  • Making machine learning models work effectively with a relatively small dataset (~420 elements).

Accomplishments we’re proud of

  • Successfully integrated mapping, interactive charts, and predictive simulations into a cohesive interface.
  • Built a functioning poverty prediction model that responds to user input scenarios.

What we learned

  • How to source and preprocess census and city data.
  • Practical experience working with GIS data and integrating it with React and Leaflet.

What’s next

  • Gathering more data to improve the accuracy and reliability of the ML model.
Share this project:

Updates