Project Story — NeuraCity: The City That Sees, Feels, and Hears

Inspiration

I have always been interested in how cities operate beneath the surface. Every moment, cities experience thousands of micro-events—accidents, noise spikes, street issues, emotional shifts in neighborhoods—yet most city systems only react slowly, if at all.

That inspired a question:

What if a city could actually see what people see, feel what communities feel, and hear what neighborhoods hear?

From that question, I built NeuraCity, a city platform designed to “perceive” and respond in real time.

About the Project

NeuraCity is a smart-city system that connects:

Seeing: Citizens upload an image of an issue, and their device’s GPS provides the exact location.
Feeling: I analyze synthetic posts using NLP to compute a city-wide emotional mood map.
Hearing: Noise data is simulated per road segment to enable quiet walking routes.
Thinking: AI generates emergency summaries, repair suggestions, and contractor recommendations.
Acting: The platform provides intelligent driving, eco, and quiet-walking routes.

All components work together to make the city more responsive and human-aware.

How I Built It

Frontend (React, Tailwind, Leaflet)

I built a clean, fast interface that supports:

A mandatory image-first reporting flow
Automatic GPS location access
Interactive maps for issues, mood, noise, and routes
A simple admin panel for emergency and work-order review

Backend (FastAPI + Supabase)

I used FastAPI to handle:

Upload processing
Severity + urgency scoring
Routing (drive, eco, quiet walk)
Noise modeling
Mood aggregation
Automatic action generation

Everything is stored in Supabase Postgres, including issues, contractors, mood scores, noise data, and work orders.

AI Systems

NeuraCity uses two AI systems:

HuggingFace sentiment model
- Processes synthetic posts
- Computes neighborhood mood scores
Google Gemini
- Generates emergency summaries for accidents
- Suggests repair materials
- Identifies contractor specialties
- Helps classify “other” issues

Synthetic Data

I generated synthetic datasets for:

Traffic
Noise
Emotional posts
Event-driven spikes

This allowed me to simulate a functioning city without real APIs.

What I Learned

Cities have emotional patterns, which can affect routing and stress levels.
Image + GPS reporting improves accuracy dramatically.
Noise matters, especially for walking comfort and sensory-friendly routing.
AI can support real infrastructure decisions, not just summarize text.
Synthetic data can be extremely realistic with the right constraints.

Challenges I Faced

Getting browser-based geolocation to behave consistently.
Designing a fast image upload/storage flow.
Building realistic noise models for routing.
Combining traffic, noise, urgency, and emotional data into a single routing engine.
Making Gemini output structured, reliable responses.

Final Reflection

NeuraCity is my attempt to imagine a city that interacts with people in a more human way. Through geospatial modeling, AI reasoning, and careful data design, I built a system where the city can: