SafeGrid: AI-Powered Proactive Safety Intelligence for Women
Link to open source: https://github.com/ashmita-web/safegrid
Link to Live Project: https://drive.google.com/file/d/1dAE5PCvxiQQO9XlF-OVdurv4rHdadYJy/view?usp=sharing
SafeGrid is a location-based personal safety platform that helps users understand nearby risks, choose comparatively safer routes, report incidents and receive timely alerts. Initially focused on Delhi, Gurugram and Noida, it combines official warnings, trusted news, community reports, historical incidents and environmental conditions to generate explainable risk maps.
The platform divides cities into H3 hexagonal cells and calculates risk using an ensemble of acute incidents, chronic patterns and environmental factors. It uses the lightweight MiniLM transformer(DistilBERT) with calibrated machine-learning models to identify relevant incidents, classify their type, estimate severity and detect duplicates.
SafeGrid includes a Flutter mobile app with safety maps, route comparison, SOS, Guardian walks, trusted contacts and check-ins. Its guarded feedback loop uses verified data to improve the model, while new models are deployed only after passing strict evaluation checks. The long-term goal is to provide scalable safety intelligence for individuals, businesses, universities, mobility platforms and public organizations.
Features
- H3-based interactive safety and risk maps
- Comparatively safer route selection
- Nearby incident and emergency alerts
- Community incident reporting
- Explainable acute, chronic and environmental risk scores
- SOS assistance and emergency actions
- Guardian mode for monitored walks
- Trusted contacts and safety Circle
- Timed safety check-ins
- AI-based incident classification and duplicate detection
- Confidence scoring and source verification
- Guarded model-learning feedback loop
- Offline mobile risk-map support
- English and Hindi language support
This build was uploaded as a hackathon project
