SafeBuddy

Inspiration

Walking alone—especially at night—can feel unsafe for many students and commuters. Most safety apps rely on manual check-ins, which aren’t always possible in a real emergency. We wanted to build something proactive: a system that notices when something is wrong, even if the user can’t reach their phone. That idea became SafeBuddy, an AI-enhanced walk-home safety companion.

What it does

SafeBuddy tracks your movement in real time while you walk. If you suddenly stop moving for too long—a potential sign of danger—the app automatically detects the anomaly and triggers an alert. The system runs in the browser, requires no installation, and uses a simple, friendly UI so anyone can use it quickly.

How we built it

We built the frontend using HTML, CSS, and JavaScript, leveraging the browser’s Geolocation API for continuous location polling. Our backend is a lightweight Flask server with endpoints for starting, updating, and ending walk sessions.

We used the Haversine formula to calculate real-world movement between GPS points and designed a simple anomaly detection system:

If user stops moving (<5 meters) for ≥60 seconds → trigger alert

This approach allowed us to build a functional safety system without requiring external datasets or heavy machine learning.

Challenges we ran into

Ensuring consistent, accurate GPS readings across different devices
Calibrating thresholds to avoid false positives
Handling high-accuracy geolocation without draining battery
Creating an intuitive user flow with minimal UI elements
Managing backend session state in a clean, lightweight way

Accomplishments that we're proud of

Built a fully functional safety tool in under 24 hours
Designed a proactive detection system instead of a manual check-in app
Created a clean, modern interface that works anywhere
Successfully implemented real-time anomaly detection logic
Produced a demo that immediately resonates with judges and users

What we learned

We learned how to handle geolocation data effectively, how to design lightweight real-time systems, and how to build safety-focused applications where clarity and reliability matter more than complexity. We also gained experience tuning thresholds for movement analysis and designing simple but meaningful UX flows.

What's next for SafeBuddy

We plan to expand SafeBuddy with several new features:
SMS or email alerts to emergency contacts
Short audio capture when anomalies occur
Route safety scoring using crime and lighting data
ML-powered fall detection using accelerometer patterns
Live map tracking for trusted contacts
AI environment understanding (e.g., detecting shouting, traffic, distress)