FINDME: AI vs HUMAN TRAFFICKING

FINDME is a global AI-powered platform designed to help locate missing and trafficked victims, making life-saving technology accessible to every corner of the world.

Comment

FINDME: Real-Time AI for Trafficked Victims and missing people

Inspiration

Human trafficking and missing persons are global crises where every minute counts. Traditional reporting systems are slow and reactive, often delaying critical interventions. We were inspired to create FINDME, a real-time AI platform that empowers communities and law enforcement to identify and locate missing individuals instantly, turning reactive processes into proactive solutions.

What it does

FINDME combines face and voice recognition to detect missing persons in real-time. Users can upload photos and audio through a web interface, while the system continuously scans live video feeds, compares embeddings against a secure database, and automatically sends SMS alerts with location details to family or authorities.
It’s a multi-modal, cloud-backed, AI-driven detection system designed for immediate, actionable results.

How we built it

Frontend: Next.js + React + Tailwind CSS for a fast, responsive user experience.

Backend: FastAPI with Python for API endpoints, image/audio processing, and Firebase integration.

AI Models: InsightFace for facial recognition, Resemblyzer for voice matching, both running on ONNX Runtime.

Database & Storage: Firebase Firestore for structured data, Firebase Storage for media.

Notifications: Sinch SMS API with geolocation for real-time alerts.

Real-time Detection: Multi-threaded video/audio pipelines, temporal smoothing, and embedding caching for speed.

ARM Optimization: Works efficiently on Raspberry Pi and edge devices without GPU.

Challenges we ran into

  • Achieving real-time face and voice recognition on ARM devices with limited processing power.
  • Efficiently managing embedding storage and comparison for multiple users simultaneously.
  • Combining face + voice matching in a reliable, seamless detection system.
  • Avoiding SMS spamming while ensuring timely alerts with geolocation.
  • Ensuring data security and privacy while storing sensitive personal information.

Accomplishments that we're proud of

  • Built a fully functional end-to-end system from reporting to real-time detection.
  • Successfully implemented multi-modal AI (face + voice), rare in hackathon projects.
  • Optimized for edge deployment, running on ARM devices with CPU-only inference.
  • Integrated real-time SMS alerts with location tracking.
  • Delivered clear, professional documentation demonstrating production-level design and security best practices.

What we learned

  • Real-time AI pipelines require careful threading, caching, and optimization to function smoothly.
  • Edge deployment on ARM devices is possible with CPU-only inference and model optimization.
  • Multi-modal recognition (face + voice) improves accuracy and reliability of detection.
  • Privacy, security, and efficient cloud storage are critical for sensitive data handling.
  • Small details like temporal smoothing and SMS cooldowns significantly improve user experience.

What's next for FINDME: AI vs HUMAN TRAFFICKING

  • Expand detection to include gait, movement patterns, and anomaly analytics.
  • Launch a mobile app for law enforcement and community volunteers.
  • Introduce AI-assisted predictive analytics to identify high-risk scenarios proactively.
  • Implement cloud-based batch processing for large-scale, city-wide monitoring.
  • Ensure global scalability, multi-language support, and strict ethical and privacy compliance.
  • Explore partnerships with NGOs and authorities to make FINDME a trusted tool against human trafficking.

Built With

Share this project:

Updates