Back to Projects

Monlogement

French property listing platform with AI-powered recommendations, scam detection, and advanced search capabilities.

View Project
Monlogement

Project Overview

Advanced property listing platform for France featuring complex search capabilities across cities, departments, postal codes, and regions. Utilizes Meilisearch for fast filtering, Mapbox for interactive maps with custom clusters, automated data importing from agency FTP systems, AI-powered recommendation system using collaborative filtering, and intelligent scam detection for imported listings.

Role

Lead Developer and AI Integration Architect

Technologies Used

LaravelTailwind CSSMeilisearchMapboxMySQLRedis Pub/SubFTPAWSFastAPIfast.ai

Tasks

  • Implemented Meilisearch for complex search and filtration of French administrative areas
  • Integrated Mapbox with custom clusters and popups for interactive property visualization
  • Imported and cleaned GeoJSON data for France area borders using GIS tools
  • Created dynamic multi-step property ad creation form with validation
  • Built custom image upload functionality with drag-and-drop interface
  • Integrated Scaleway TEM API and SMTP for reliable email delivery
  • Implemented queued jobs for asynchronous email sending
  • Connected to Digital Ocean Spaces for scalable image storage
  • Deployed through Laravel Forge with custom server configuration
  • Created bash scripts for database backup and restoration automation
  • Added Meilisearch as system service in server for persistent operation
  • Developed FastAPI microservice for AI-powered recommendation engine
  • Implemented collaborative filtering with content-based metadata using fast.ai for personalized property recommendations
  • Built automated scam detection system analyzing FTP imported ads
  • Established Redis Pub/Sub communication between Laravel, FTP import service, and FastAPI
  • Created machine learning pipeline for detecting fraudulent listings patterns

Achievements

  • Implemented Redis caching and Pub/Sub improving performance significantly
  • Optimized database queries avoiding N+1 lazy loading issues across the platform
  • Created automated alert system based on user search criteria with cron jobs
  • Optimized image sizes on upload reducing storage costs by 60%
  • Implemented FTP ads importing from agency file systems using optimized jobs with middleware
  • Improved query performance using Laravel Debug Bar and Telescope
  • Deployed AI recommendation system increasing user engagement by 45%
  • Achieved 92% accuracy in scam detection protecting users from fraudulent listings
  • Built real-time communication architecture using Redis Pub/Sub between 3 services
  • Reduced manual moderation workload by 70% through automated scam detection
  • Implemented collaborative filtering recommending properties with 35% click-through rate

Contributions

  • Implemented DTO pattern for input transfer and sanitization across services
  • Created global helper functions autoloaded through Composer for code reusability
  • Built Services layer for shared logic between web and API endpoints
  • Used model event observers for cleaner code architecture and separation of concerns
  • Implemented API resources and form requests for custom validations
  • Managed user preferences and filter states with sessions for personalized experience
  • Refactored code and restructured filesystem for better maintainability
  • Architected microservices ecosystem with Laravel (main app), Python FTP importer, and FastAPI AI service
  • Designed Redis Pub/Sub messaging patterns for asynchronous service communication
  • Built collaborative filtering recommendation model using fast.ai analyzing user interactions
  • Developed scam detection ML model identifying fraudulent patterns in property listings
  • Created data preprocessing pipeline for training recommendation and fraud detection models
  • Implemented real-time property recommendations based on user behavior and preferences
  • Built automated flagging system for suspicious ads from FTP imports with confidence scores
  • Integrated FastAPI service handling both recommendations and scam detection with shared infrastructure
  • Established monitoring and logging for microservices communication and AI model performance