City Health Observatory

About City Health Observatory

Inspiration

Smart cities rely on interconnected systems—traffic, air quality, power, water, and noise. But there’s no simple way to understand overall city health or how failures spread across regions.

We built this for Luddy Hackathon Case 2 to solve that gap: a real-time Smart City Health Leaderboard that continuously scores districts and visualizes cascading failures.

Urban Health Score (UHS)

Each district’s health is computed using weighted sensor signals:

[ \mathrm{UHS} = \sum_i w_i s_i ]

To reduce noise, we apply temporal smoothing:

[ \mathrm{UHS}{new} = (1-\alpha),\mathrm{UHS}{window} + \alpha,\mathrm{UHS}_{prev} ]

What it does

A real-time leaderboard that scores and ranks city health every few seconds.

Streams sensor-like data and computes a 0–100 health score
Displays a live ranked leaderboard
Detects cascading failures across neighboring districts
Applies geographic ripple penalties over time

Key idea: A failure in one district doesn’t stay isolated—it spreads.

The system supports:

Local mode: SQLite, no broker
Streaming mode: Redpanda + Flink / Python worker + Redis + PostgreSQL

How we built it

Backend: Python 3.12, FastAPI, Pydantic
Streaming: Redpanda (Kafka-compatible), PyFlink or Python worker
Storage: Redis (real-time), PostgreSQL / SQLite (history)
Frontend: HTML/CSS/JS dashboard
Infra: Docker + Compose (modular profiles)

Challenges we ran into

Flink setup issues: PyFlink + JVM caused inconsistent builds across machines
System coordination: Syncing broker → processor → Redis → API required careful orchestration
Dual-mode complexity: Maintaining both local and streaming modes without breaking the UI
Data realism: Ensuring synthetic data still tells a believable story