ClimateShield: AI Weather Relief Finder

Know where to go when the weather turns: live weather, the nearest places that help for how you travel, and a personal AI safety plan.

Comment

Inspiration

Sudden weather catches people off guard. I imagined someone walking in 42°C heat who feels dizzy, or someone who gets a "rain in 5 minutes" alert with no umbrella and no idea where to shelter. Weather apps show numbers and maps show everything, but nothing tells you where to go right now. I wanted to build that missing link.

What it does

ClimateShield is an AI-assisted recommendation system for extreme weather. It takes your location, the live weather there, and how you are travelling, and recommends the nearest places that really help plus a personal safety plan.

ClimateShield reads the live weather for your location and shows the nearest places that really help, like water, shelter, pharmacies, covered bus stops, fuel stations or umbrella shops.

  • It changes what it shows by weather and by how you travel: walk, bicycle, motorbike, car or bus.
  • It shows the distance and travel time for each place, and the route in one tap.
  • Need something now? shows the single nearest water point, pharmacy or hospital.
  • An AI advisor writes a simple safety plan (what to do and when) and answers questions like "I feel dizzy from the heat".
  • The whole app changes its look with the weather: sunny, rain, storm, snow, fog, heatwave, cold wave and night. You can also pick any weather from a menu to preview it or plan ahead.

How I built it

  • Front end: Next.js, React, TypeScript and Leaflet.
  • Data: Open-Meteo for weather and air quality, OpenStreetMap for places and routes.
  • AI: Gemini writes the safety plan and the chat answers.
  • Backend: small server routes on Vercel keep the AI key private and fetch map data.
  • Safe AI: the AI only sees live weather, the travel mode and the real nearby places. Its place suggestions are checked against that list, so it can't make up places.

Challenges I ran into

  • Slow AI: my first model took up to a minute to reply, so I switched to a faster one that answers in a few seconds.
  • Unreliable map data: the free map servers were often busy or blocked browsers, so I moved the requests to my own server, added backups and saved results.
  • AI making things up: I made sure it can only recommend real places from the list.
  • Different travel modes: a walker and a motorbike rider need very different help, so places and times change for each mode.

Accomplishments that I'm proud of

  • It works live on any phone with no sign-up.
  • The whole app changes with the weather, so it feels alive.
  • The AI is based on real data and real places, not guesses.
  • It gives useful first-aid style guidance, with a clear note to call emergency services when needed.
  • I finished a complete working product on my own.

What I learned

AI is most useful when it is given real, live data and clear limits. I also learned to plan for slow or failing services, like free map servers and AI limits, so the app still works when something is busy.

What's next for ClimateShield: AI Weather Relief Finder

  • Let people add and review places, so the map gets better over time.
  • Send alerts before rain or heat reaches saved places.
  • Add more languages.
  • Show if a place is open right now.

Built With

Share this project:

Updates

Submission history