DatingApp - AI-Powered Smart Dating Platform

This project is a high-performance, full-stack web application developed with a .NET 10 backend and an Angular 21 frontend.

It serves as a comprehensive case study in building modern, AI-integrated enterprise software. Beyond standard CRUD operations, it implements a Hybrid RAG (Retrieval-Augmented Generation) architecture, showcasing how to bridge the gap between structured relational data (SQL Server) and unstructured semantic data (Vector Databases).

Live Demo: https://dating-2025.azurewebsites.net/ (Note: The free-tier Azure App Service may experience a cold start.)

Application Preview

Member listing with filtering options.

🧠 AI & Hybrid RAG Architecture

The project demonstrates a production-ready AI Engineer stack integrated within a Clean Architecture (4-project structure):

1. Semantic Matchmaking

   • Vectorization: User profiles are transformed into high-dimensional vectors. While the initial version used local ONNX models, the system was refactored to use the Hugging Face Inference API.
   • The Model: Uses the all-mpnet-base-v2 model ($768$ dimensions). This choice provides superior semantic accuracy compared to smaller models while maintaining a serverless, lightweight backend footprint.
   • Vector Database: Embeddings are stored in Qdrant. The system performs Cosine Similarity searches to match users based on the "intent" of their descriptions rather than simple keywords.

2. Cerebras-Powered Chat Intelligence

   • Near-Zero Latency: Integration with Cerebras Systems (via gpt-oss-120b) ensures near-instant AI responses.
   • Intelligent Advice: The Chat Assistant uses Microsoft.Agents.AI to provide context-aware ice-breakers and dating advice based on chat history.

3. Dual-Store Synchronization (SQL + Qdrant)

   • The application maintains state consistency between SQL Server (structured user data) and Qdrant (semantic data).
   • Automatic Sync: A robust synchronization logic is implemented during the database seeding process, ensuring that the vector space is always a reflection of the relational database.

🌍 Geospatial Intelligence & Proximity Search

Beyond AI matching, the platform implements high-precision location-based filtering to ensure relevant local connections:

• Spatial Data Processing: Integrates NetTopologySuite to handle complex GIS (Geographic Information System) data. User locations are stored as Point types using the SQL Server Geography data type.
• Automated Geocoding: Leverages the OpenCage API to convert plain-text city and country data into precise GPS coordinates during user registration and profile updates.
• Performant Proximity Filtering: Uses specialized spatial indexing in SQL Server to perform distance-based queries (e.g., "Find matches within 50km") with near-zero latency, avoiding expensive row-by-row calculations.
• GeoJSON Integration: Custom JSON converters ensure seamless communication between the .NET backend and the Angular frontend by following the standard GeoJSON format.

🛠 Technology Stack

Backend (.NET 10)

• Architecture: Clean Architecture with clear separation of concerns.
• Standardized AI: Built using Microsoft.Extensions.AI (MS Agent Framework) for provider-agnostic AI integration.
• Spatial Intelligence: Uses NetTopologySuite for physical distance calculations and OpenCage API for geocoding.
• Security: JWT-based authentication with ASP.NET Core Identity.
• Real-time: SignalR for presence tracking and instant messaging.

Frontend (Angular 21)

• State Management: Leveraging the latest Angular Signals for reactive and efficient UI updates.
• Modern Styling: Built with Tailwind CSS and DaisyUI for a clean, responsive user experience.
• Image Handling: Cloudinary integration for optimized cloud-based image transformations.

🚀 Local Development Setup

1. Infrastructure (Docker) Ensure Docker Desktop is running. Start the vector database:

   docker compose up -d qdrant
   

2. Configuration Update appsettings.Development.json with your API keys:

   {
     "HuggingFace": {
       "ApiKey": "your_hf_token",
       "ModelId": "sentence-transformers/all-mpnet-base-v2"
     },
     "CerebrasSettings": {
       "ApiKey": "your_cerebras_key"
     },
     "ConnectionStrings": {
       "DefaultConnection": "Server=YOUR_SERVER;Database=datingdb;Trusted_Connection=True;"
     }
   }
   

3. Run Application

   # Run Backend
   cd DatingApp.Presentation && dotnet run
   
   # Run Frontend (in a separate terminal)
   cd client && npm start
   

Project Purpose

This repository is designed to demonstrate proficiency in:

• Integrating Generative AI and Vector Search into enterprise workflows.
• Implementing Clean Architecture in a .NET 10 environment.
• Managing complex infrastructure (Docker, Azure, Vector DBs).
• Building high-performance, reactive UIs with the latest Angular features.