---

🚀 Headstarter Fellowship Hackathon - Ruby Track (hackathon finalist)

Project: AI-Powered Complaint Categorization System for Spend Ruby 💼

📝 Overview

This project is part of the Ruby Track Hackathon in the Headstarter Fellowship. The hackathon features 7 companies, each looking to hire top talent. As part of this challenge, our team has chosen to work with Spend Ruby, a financial platform designed for modern businesses. The goal of this project is to develop an AI-powered system that categorizes consumer complaints, providing meaningful insights and improving customer service.

💡 Project Description

We are developing an AI-powered complaint categorization system that processes text-based consumer complaints. The system will:

1. Process Consumer Complaints: Analyze text, audio, and image based complaints submitted by consumers.
2. Categorize Complaints: Assign appropriate product and sub-product categories to each complaint.
3. Store Categorized Data: Saved the categorized complaints in a database and vector database for RAG pipeline.
4. Enhance with a RAG Pipeline: Implemented a Retrieval-Augmented Generation (RAG) pipeline using a vector database to retrieve and compare related complaints based on similarity.

🔑 Key Features

• Text Processing: Utilize natural language processing (NLP) techniques to understand and process consumer complaints.
• Categorization: Employ machine learning models to accurately categorize complaints into predefined product and sub-product categories.
• Database Integration: Store the categorized data in a robust database system for easy access and management.
• RAG Pipeline: Use a RAG pipeline to improve the retrieval of similar complaints, aiding in better analysis and resolution.

🛠️ Technologies Used

Overall architecture diagram:

• Frontend:

  • HTML for user input forms and app structure.
  • Tailwind CSS framework for rapidly styling and customizing our app.
  • Shadcn Components to easily tweak styles, themes, and behaviors.

• Backend:

  • NextJS for rapid routing API development.
  • Supabase PostgreSQL built-in database for relational data, and for the vector database
  • LangChain RAG Pipeline for integration with Pinecone to enable retrieval of similar complaints.
  • Google Gemini API for complaint categorization and summary generation.
  • Hugging face Whisper model for speech to text
  • OCR Space for image to text

👥 Team Members

• Jinay Patel - GitHub
• Abdelghani Bensaïh - GitHub
• Nebila Wako - GitHub

📞 Contact

For any inquiries or further information, please feel free to contact us.

---