RecceLab's LLM-powered Dashboard
A sophisticated full-stack marketing analytics platform with LLM-powered insights and machine learning predictions for marketing campaign data.
A modern full-stack marketing analytics platform that leverages LLM technology and machine learning to provide advanced insights and predictions on marketing campaign performance across different channels, demographics, and regions.
Main dashboard interface showing key marketing metrics and performance indicators
Architecture Overview
Built with a microservices architecture, the application consists of:
- Frontend: Next.js 15 with TypeScript, React, and Tailwind CSS
- Backend API: Python Flask REST API for data management
- LLM Backend: Specialized Flask API for AI-powered analytics
- Database: MongoDB for flexible and scalable data storage
- Nginx: Reverse proxy for efficient request routing
- Docker: Containerization for consistent deployment
Secure login interface with modern design
User management modal for admin controls
Key Features
- Interactive Marketing Dashboard: Comprehensive analytics with ROI metrics, channel performance, and demographic segmentation
- LLM-Powered Analysis: Natural language query processing to generate charts, descriptions, and reports automatically
- Machine Learning Predictions: Prophet time-series forecasting for revenue and ad spend projections
- Multi-dimensional Data Visualization: Analysis by marketing channel, age group, and country with interactive filters
- Data Management: Database uploader/editor for marketing campaign data with validation
Revenue and ad spend visualization with performance metrics
Heatmap showing performance across different marketing channels
Technical Challenges
- LLM Integration: Designed a modular pipeline system for classifying queries (chart, description, report), selecting appropriate data collections, and generating visualization or text responses
- Asynchronous Processing: Implemented job tracking and status management for long-running LLM queries to prevent timeout issues
- Real-time Data Visualization: Created state-managed, filtered views for different segmentation perspectives while maintaining performance
- ML Prediction Pipeline: Integrated Prophet forecasting with visualization components for comparing actual vs. predicted metrics
The LLM-powered report builder interface for generating insights from campaign data
Implementation Highlights
- Data Visualization Architecture: Designed specialized chart components using Recharts for channel contribution, cost metrics heatmap, revenue analysis, and ML predictions
- LLM Query Pipeline: Created a robust classification and processing system with automatic cleanup for memory management
- Context Providers: Implemented React context for optimized state management across components
- Microservices Communication: Established efficient API patterns between frontend, main backend, and LLM service