RadonWatch

beautiful UI
diagrams and educational content
final quiz
test your home with an interactive map
real 3d visualization of your results
compare your results with real data

Inspiration

Radon is the second leading cause of lung cancer in Canada, responsible for ~3,300 deaths annually. Yet most Canadians have never heard of it. 1 in 5 Canadian homes exceeds Health Canada's safety guideline of 200 Bq/m³, but testing rates remain dismally low because people simply don't know about the risk.

There is basically a major, preventable health crisis hiding in plain sight. The problem isn't just awareness, it's that radon education is dense, scientific, and intimidating. So I thought what if we could make learning about radon more modern as most eductional resources are boring and hard to read?

RadonWatch was created from the belief that education drives action. If people understand what radon is, where it comes from, and how it affects them personally, they'll test their homes and potentially save their lives.

What it does

To put it simply: RadonVision is an over-engineered learning tool that combines education with predictive technology:

Interactive Educational Journey

Progressive lessons unlock as users learn
Topics: Uranium decay chains, health effects, mitigation strategies
Gamified quizzes reinforce learning
Visual explanations of complex nuclear physics

Risk Prediction System

Users input home characteristics (region, foundation type, age, soil geology, ventilation)
A heuristic model calculates personalized radon risk
Simulated "neural network training" animation shows the system "learning"
Results displayed with confidence intervals and regional comparisons

3D Radon Visualization

Real-time particle simulation showing radon rising from soil through foundation
Interactive controls: adjust air pressure, toggle ventilation systems
Color-coded danger zones based on concentration
Built with Three.js for stunning visual impact

Canada-Wide Risk Dashboard

Real Health Canada data visualized on interactive map
Regional statistics and trends
Comparison charts: user's home vs. neighbors vs. national average
Actionable next steps and mitigation cost estimates

How we built it

Tech Stack:

Frontend: Next.js 14 (React), Tailwind CSS v4
3D Graphics: Three.js + React Three Fiber
Data Viz: Recharts for charts and graphs
Data Source: Health Canada's Cross-Canada Radon Survey (14,000+ home measurements)
Deployment: Vercel

Here's where the overengineering comes in. Ididn't have time to train a real ML model in 10 hours, so I built something better for education: a research-backed heuristic model disguised as AI. Our prediction algorithm uses weighted factors based on peer-reviewed radon literature:

Our prediction algorithm uses weighted factors based on peer-reviewed radon literature:

RadonRisk = BaseLevel × RegionMultiplier × AgeMultiplier 
            + FoundationAdder 
            × SoilTypeMultiplier 
            × VentilationMultiplier 
            + RandomVariance

Why this works:

Scientifically accurate: Factors derived from actual radon research
Fast to implement: No training data needed
Educational value: Shows users which factors matter most through feature importance charts
Convincing presentation: Fake training animations with loss curves, accuracy metrics, epoch counts
Transparent: We're upfront that it's educational, not diagnostic

The model incorporates:

Geographic data: Interior BC has 2.5x higher risk than coastal areas
Building age: Newer homes (2010+) have 1.4x multiplier due to energy efficiency trapping radon
Foundation types: Basements add +40 Bq/m³ base risk
Soil geology: Granite regions have 1.5x multiplier
Ventilation quality: Poor ventilation increases risk by 1.3x

We validate predictions against Health Canada's dataset to ensure realism.