Inspiration
Air quality is something most people only think about when it's visibly bad — smoke, smog, or a stuffy room. We wanted to make it visible and actionable before it gets that bad, especially indoors and in urban areas where PM2.5 and other pollutants can spike unnoticed. AeroSense was born out of wanting a simple, live dashboard that tells you not just "what's the air like right now" but predicts when it's heading toward hazardous levels.
What it does
AeroSense is an AI-based air quality prediction and monitoring system with a live dashboard. It tracks six key metrics in real time:
- PM2.5 and PM10 (particulate matter)
- CO (carbon monoxide)
- NO2 (nitrogen dioxide)
- Temperature and humidity
Based on these readings, it classifies current air quality (e.g. Good, Moderate, Hazardous) and flags warnings when pollution levels are predicted to become dangerous, so users can take precautions in advance rather than reacting after the fact.
How we built it
- Backend: [Flask / Python — confirm], serving live sensor/simulated data through API endpoints
- Frontend: A lightweight live dashboard that polls the backend every few seconds and updates metrics and a real-time PM2.5 trend chart
- Prediction logic: [Rule-based thresholds / ML model — confirm which] to classify air quality and generate hazard warnings
- Deployment: Hosted live on Render at aerosense-9o0o.onrender.com
Challenges we ran into
- Getting live data to update smoothly on the frontend without lag
- Deploying and configuring the build/start commands correctly on Render
- [Add: sensor calibration / noisy readings / data smoothing, if relevant]
Accomplishments that we're proud of
- A fully working, publicly deployed live dashboard
- Real-time predictive air quality classification, not just raw sensor display
What we learned
- [Add: e.g. working with Flask APIs, deploying to Render, handling live data polling, teamwork under hackathon time pressure]
What's next for Aerosense
- Integrating real hardware sensors for live field deployment
- Improving prediction accuracy with a trained ML model
- Adding historical trend analysis and location-based comparisons
Log in or sign up for Devpost to join the conversation.