AQIntel – AI-Powered Hyperlocal Air Quality Intelligence

Inspiration

Air pollution is something we experience every day, but understanding what is happening at a local level is still difficult. A city may have only a limited number of monitoring stations, while pollution can change from one street or neighborhood to another because of traffic, industries, construction, weather, and wind.

This made us think: What if we could build a system that does more than simply show the AQI?

We wanted AQIntel to help people understand where pollution is getting worse, what might be causing it, what could happen next, and where attention is needed most. That idea became the starting point for AQIntel – AI-Powered Hyperlocal Air Quality Intelligence.

What it does

AQIntel brings together different sources of environmental information, including IoT sensors, government monitoring stations, weather data, satellite observations, GIS information, roads, traffic context, and industrial locations.

Instead of looking at these sources separately, AQIntel combines them to build a more detailed picture of air quality at the local level.

The system is designed to:

Monitor and understand local air-quality conditions
Identify potential pollution hotspots
Forecast future air-quality conditions
Analyze possible pollution sources
Explain why an AI model made a particular prediction
Assess environmental risk
Provide information that can support better decisions

Our goal is to move from simply asking “What is the air quality?” to asking “Why is it changing, what might happen next, and what should we pay attention to?”

How we built it

We designed AQIntel as a modular system so that different data sources and technologies can work together.

The overall pipeline follows:

Data Collection → Data Validation → Data Fusion → Hyperlocal Analysis → Forecasting → Source Attribution → Explainable AI → Decision Support

We use environmental and geospatial data to understand local conditions, machine-learning techniques for prediction, and explainability methods such as SHAP to make the AI output easier to understand.

The idea was not to build just another dashboard. We wanted to connect the data, AI, and decision-making parts into one system.

Challenges we ran into

One of the hardest parts was dealing with real-world environmental data. Different sources have different formats, timestamps, locations, resolutions, and levels of reliability. Some data can also be missing or noisy.

Another challenge was combining all these sources without losing their meaning. We had to think carefully about how weather, geography, traffic, sensors, and other environmental factors relate to each other.

We also wanted the system to be understandable. A prediction is much more useful when users can see the reasoning behind it, so explainability became an important part of our design.

Finally, we wanted AQIntel to be useful beyond a single location. Designing it in a way that can adapt to different cities, datasets, and sensors was another important challenge.

Accomplishments that we're proud of

We are proud that AQIntel brings several different areas together in one framework:

IoT-based environmental sensing
Multi-source environmental data fusion
Hyperlocal air-quality intelligence
AI-based forecasting
Pollution source attribution
Explainable AI
Geospatial analysis
Environmental risk assessment
Decision support

Most importantly, we are proud of the idea behind the project. AQIntel is not designed only to display pollution numbers. It is designed to help turn those numbers into information that people and authorities can actually understand and use.

What we learned

Building AQIntel taught us that solving a real-world problem with AI is not just about choosing the best model.

We learned how important data quality, spatial context, time, uncertainty, and explainability are when working with environmental information.

We also learned that an accurate prediction is not always enough. If people cannot understand the result or know what to do with it, the value of the prediction is limited.

Most importantly, we learned to think about AI from the perspective of the problem and the decision that needs to be made, rather than only from the perspective of the technology.

What's next for AQIntel – AI-Powered Hyperlocal Air Quality Intelligence

AQIntel is only the beginning.

Our next goal is to strengthen real-world validation and improve the system's ability to work across different locations and environmental conditions.

In the future, we would like to explore: