HeatShield - Qualcomm Challenge

Inspiration

Barcelona is facing more frequent and intense heatwaves, and extreme heat has already caused deaths, especially among older and vulnerable people. Climate shelters are essential, but in critical moments staff may not detect a collapse immediately. That urgency inspired HeatShield.

What it does

HeatShield adds a real-time safety layer inside climate shelters. It detects possible fainting/fall events related to heat stress and automatically triggers an emergency alert so staff can react faster.

How we built it

We built HeatShield using Arduino + Modulino sensors to capture environmental context, and Edge Impulse to train and fine-tune AI models optimized for Edge AI devices. On top of that, we implemented a real-time backend and dashboard that orchestrate detection events and instant emergency alerts.

Challenges we ran into

Our main challenge was balancing high sensitivity (not missing real fainting events) with low false alarms. We also had to keep latency low and ensure the system stayed reliable on edge hardware.

Accomplishments that we're proud of

We delivered an end-to-end prototype that combines sensors, edge AI inference, and automatic emergency activation in real time. We’re proud it addresses a real public-health problem with a practical, deployable approach.

What we learned

We learned that for heat-risk scenarios, reliability and response time matter more than complex features. A fast, trustworthy alert can make a real difference when every second counts.

What's next for HeatShield

Next, we want to improve robustness in crowded shelter conditions, expand training data for heat-stress scenarios, and pilot HeatShield in real shelter operations to validate impact.

Built With

• arduino
• edge-impulse
• fomo
• javascript
• python