🛡️ SafeChild: AI-Driven Personal & Child Safety Ecosystem
Empowering Communities with Real-Time Tracking, Hands-Free Voice SOS, and Bilingual AI Emergency Guidance.
🚀 Inspiration
In critical emergency situations—whether facing harassment, accidents, or distress—every second counts, yet traditional emergency response is often delayed or hindered by panic. Inspired by the urgent need for a reliable, frictionless safety net in South Asia, SafeChild was created to bridge the critical gap between panic and immediate assistance. By combining instant voice activation, automated emergency contact chains, and intelligent AI-driven first aid, SafeChild ensures that help is never more than a voice command or single tap away.
⚡ What It Does
SafeChild is a comprehensive, AI-powered safety web and mobile application engineered for speed, reliability, and ease of use in high-stress environments:
- Hands-Free Voice SOS: Instantly triggers emergency protocols via voice commands (e.g., "help help" or "বাঁচাও বাঁচাও"), even when the phone is locked.
- Automated Emergency Call & SMS Chain: Sequentially alerts trusted contacts and local emergency services (e.g., 999), dispatching live GPS tracking map links to unanswered contacts.
- Live Family & Safety Map: Built with Leaflet and OpenStreetMap for real-time location sharing, featuring secure 4-digit pairing and instant Safe/Unsafe status updates.
- Bilingual First Aid AI Assistant: Powered by Google Gemini to deliver immediate, calm, and accurate medical first-aid instructions in English, Bangla, and Banglish.
- Offline Resilience: Core emergency triggers and status configurations function seamlessly without internet connectivity via automated SMS fallbacks.
🛠️ How We Built It
SafeChild is built using a robust, production-grade modern tech stack:
- Frontend / Web Prototype: React, TypeScript, Vite, Tailwind CSS, Leaflet.js.
- Backend & Cloud Architecture: Firebase (Authentication, Firestore, Cloud Messaging), TypeScript Cloud Functions.
- AI & Intelligence: Google Gemini API integrated securely via serverless backend endpoints to ensure zero API key exposure.
- Mobile Core (In Progress): Kotlin, Jetpack Compose for native Android background services and foreground safety loops.
🧗 Challenges We Ran Into
- Android Background Restrictions: Overcoming aggressive battery optimization policies on manufacturers like Xiaomi, Oppo, and Vivo to maintain reliable background voice keyword detection.
- Bilingual AI Guardrails: Structuring context prompts for low-resource languages (Bangla/Banglish) to eliminate hallucinations and prioritize emergency medical accuracy.
- State Synchronization: Handling real-time location syncing and secure 4-digit pairing across decentralized users while maintaining strict data privacy.
🏆 Accomplishments That We're Proud Of
- National Recognition: Shortlisted among the top 118 out of 355 competing projects at the BIIN Awards 2026.
- Global Validation: Achieved finalist status in the USAII Global AI Hackathon.
- Functional Prototype: Successfully deployed a fully working web prototype featuring live GPS mapping and bilingual AI assistance.
📚 What We Learned
- The critical nuances of mobile operating system background limitations for safety-critical applications.
- Advanced prompt engineering techniques for bilingual AI deployment in emergency scenarios.
- The importance of fault-tolerant architecture (offline-first SMS fallbacks) when internet connectivity fails.
🔮 What's Next for SafeChild
- Native Android Deployment: Complete real-device testing for the Kotlin/Jetpack Compose Android app across diverse hardware.
- Global Emergency Integration: Expand the emergency number database to cover 246+ countries with local authority APIs.
- Predictive Safety Analytics: Introduce community-driven safety heatmaps and AI-powered hazard alerts to prevent incidents before they occur.
Built With
- app
- geminiapi
- leaftleft
- react
- typescript
- versel
- vite

