Overview
This project was developed as a comprehensive solution for modern social platforms, focusing on two critical aspects: efficient matchmaking and robust AI-driven scam detection. It leverages machine learning to analyze user behavior and content in real-time, significantly reducing fraudulent activities.
Architecture & Tech Stack
- Backend: ASP.NET Core 8 with WebSocket server for real-time interactions.
- AI Service: Python (FastAPI) hosting ML models (XGBoost, Sentence Transformers).
- Frontend: ReactJS (Web) and React Native (Mobile) with Tailwind CSS.
- Infrastructure: WebRTC for video calls, Redis for caching, MongoDB for flexible data storage.
- Deployment: containerized with Docker and orchestrated with K3s on a VPS infrastructure.
Key Features
1. Smart Matchmaking
Utilizes advanced algorithms to connect users based on shared interests, behavior patterns, and proximity.
2. AI Scam Detection
An independent Python service that constanty monitors conversations and profile updates using NLP and XGBoost models to identify and flag potential scammers before they can act.
3. Real-time Communication
Integrated high-quality video calling via WebRTC and instant messaging via WebSocket, ensuring low-latency communication.
4. Enterprise-Grade DevOps
Robust CI/CD pipeline using GitHub Actions, automated health monitoring, and secure tunnel deployment via Cloudflare.
My Role (System & DevOps)
- Architected the System Infrastructure and deployment strategy.
- Configured and managed the K3s/Docker environment.
- Integrated Cloudflare Tunnel for secure service exposure.
- Optimized the Service Mesh to handle gRPC and WebSocket traffic simultaneously.
🚀 Repository: github.com/huynartLZ/DATN2026-STT43