AirSentinel

Inspiration

Air pollution is no longer just a city problem—it’s a location problem. People living only a few kilometers apart often breathe very different air, yet most have no reliable way to know it. Existing air quality systems depend heavily on ground monitoring stations, which are expensive, sparse, and largely limited to major cities. This leaves rural areas, highways, and many urban neighborhoods without timely air quality information—especially during short-term pollution spikes.

We were driven by a simple question: *Why should access to air quality information depend on whether a sensor exists nearby?

What it does

AirSentinel* measures and forecasts short-term air pollutant concentrations for any location using satellite and reanalysis data. It provides:

• Location-specific pollutant estimates without relying on ground sensors
• Short-term air quality forecasts
• Health-focused insights that translate data into actionable guidance

The platform works even in unmonitored and remote regions, making air quality awareness more accessible and inclusive

How we built it

We fused satellite observations of atmospheric pollutants with reanalysis meteorological data such as wind, temperature, and humidity. These inputs are processed and aligned spatially and temporally, then fed into a machine learning–based prediction pipeline designed for short-term estimation. The results are visualized through an intuitive interface that emphasizes clarity, confidence levels, and health relevance

Challenges we ran into

• Aligning satellite and reanalysis data with different resolutions
• Handling uncertainty in regions without ground validation
• Balancing model complexity with explainability
• Designing outputs that are both scientifically sound and easy to understand

Accomplishments that we're proud of

• Built a sensor-independent air quality system
• Enabled air pollution estimation for any location, including rural areas
• Focused on short-term, actionable insights rather than static AQI values
• Created a solution that is scalable and aligned with real-world public needs

What we learned

• Data fusion significantly improves short-term air quality estimation
• Simpler, explainable models are more suitable for public-facing systems
• Translating technical data into health-aware insights is critical for impact
• Responsible AI means acknowledging uncertainty, not hiding it

What's next for AirSentinel

• Expand validation using more ground monitoring data
• Integrate low-cost sensors where available to improve accuracy
• Add real-time alerts for vulnerable populations
• Deploy the platform at a regional or city level for pilot testing

Built With

• api
• gemini