SheGuardAI

Inspiration

Every minute counts during an emergency, yet many victims are unable to unlock their phones or manually trigger an SOS. We envisioned a solution where the smartphone itself becomes an intelligent guardian—capable of recognizing danger, understanding context, and responding instantly without waiting for user interaction.

What it does

SheGuardAI is an AI-powered personal safety assistant that continuously monitors audio, body motion, and GPS location using a smartphone's built-in sensors. By combining these signals, it accurately detects dangerous situations while minimizing false alarms. Once a threat is confirmed, it automatically initiates emergency calls, shares live location, records evidence, and activates visual and audio deterrents.

How we built it

We developed SheGuardAI using a multimodal AI pipeline that combines: Audio Analysis with Librosa feature extraction and a Random Forest classifier. Motion Detection using smartphone accelerometer data to identify falls, violent movements, and abnormal activity. Location Intelligence to detect movement in high-risk zones. Risk Fusion Engine that merges all sensor outputs into a single confidence score. Native Android Integration for automatic calling, siren activation, live location sharing, and emergency recording. Everything runs efficiently on a smartphone without requiring additional hardware.

Challenges we ran into

Reducing false alarms while maintaining fast emergency detection. Synchronizing multiple sensor inputs in real time. Optimizing AI models for low battery consumption. Building a seamless emergency response workflow using native smartphone capabilities. Ensuring user privacy by processing sensitive data securely.

Accomplishments that we're proud of

Built a fully hardware-independent safety solution using only smartphone sensors. Designed a multimodal AI risk engine for accurate threat detection. Enabled automatic emergency response without requiring user interaction. Created a privacy-first architecture with efficient on-device processing. Developed a scalable solution that can benefit millions of smartphone users.

What we learned

This project taught us that combining multiple sensor signals significantly improves AI reliability. We gained hands-on experience in multimodal machine learning, sensor fusion, Android hardware integration, real-time AI optimization, and designing technology that directly impacts human safety.

What's next for SheGuardAI

Our next goal is to make SheGuardAI even smarter by integrating: Offline Edge AI for faster detection. Smartwatch and wearable integration. Personalized risk prediction using user behavior patterns. AI-powered voice distress understanding in multiple languages. Direct integration with emergency services and trusted community responders. Global deployment as an accessible personal safety platform.

Built With

• flutter
• geo-locator
• google-maps
• machine-learning
• motion-sensor
• mycaption-speech-to-text
• python
• sensor+
• siren
• twilio