Nearwise

AI-powered local discovery that understands natural language and ranks nearby places with Google Places and OpenAI.

Comment

Inspiration

Finding good places in an unfamiliar city is still harder than it should be. Search results are often keyword-based and require users to browse through many options themselves. We wanted to build an app that understands natural language requests and returns a ranked list of places that actually matches the user's intent.

What it does

Nearwise is an AI-powered local discovery app for iOS.

Users can ask questions like:

  • "Best brunch near me"
  • "Cheap eats tonight"
  • "Hidden coffee shops locals love"

Nearwise combines Google Places with GPT-5.6 to understand the request, evaluate nearby candidates, and return an intelligent ranking instead of a simple keyword search.

Users can also view detailed place information and save favorite locations for later.

How we built it

  • SwiftUI
  • Google Places API
  • OpenAI GPT-5.6 Responses API
  • Vercel Serverless Functions
  • GitHub
  • Xcode

The iOS app retrieves nearby candidates from Google Places, while the backend uses GPT-5.6 to analyze the user's request, compare candidates, and generate an explainable ranking before returning the results to the app.

Codex was used extensively throughout development for architecture planning, implementation, debugging, refactoring, testing, and deployment.

Challenges we ran into

The biggest challenge was making AI ranking reliable and responsive enough for production use.

We improved structured outputs, reduced latency, added retry logic, optimized candidate selection, and built deterministic fallbacks so the app remains usable even when AI responses are delayed.

We also spent significant time polishing the user experience and ensuring smooth interaction between the iOS client, Google Places, and the backend.

Accomplishments that we're proud of

  • End-to-end AI-powered local search
  • Natural language place recommendations
  • Production deployment on Vercel
  • Responsive SwiftUI interface
  • Reliable ranking pipeline with graceful fallback handling

What we learned

This project taught us how to combine LLM reasoning with real-world location data while balancing latency, reliability, and user experience.

We also learned how valuable AI-assisted development can be when combined with careful testing and iterative refinement.

What's next for Nearwise

Future plans include:

  • Multi-city support
  • Personalized recommendations
  • Trip planning
  • Better ranking personalization
  • Additional map and navigation features## Inspiration

Built With

Share this project:

Updates