AROL Group Smart Chatbot

An AI-powered chatbot designed to provide accurate, domain-specific responses about AROL Group’s products and services. The chatbot integrates LLM fine-tuning, retrieval-augmented generation (RAG), and web crawling to enhance response quality and domain-specific knowledge.

📌 GitHub Repository
📌 Chatbot Demo on Hugging Face


🔹 Key Features

  • Web Crawling & Data Scraping → Extracted structured content from AROL Group’s websites using Scrapy and BeautifulSoup.
  • Data Preprocessing → Converted raw data into a Q&A format for training.
  • LLM Fine-Tuning → Adapted LLaMA 3.2 (3B parameters) using LoRA (Low-Rank Adaptation) for domain-specific responses.
  • Retrieval-Augmented Generation (RAG) → Integrated Pinecone for efficient vector-based retrieval.
  • Deployment → Hosted on Hugging Face Spaces with Gradio for real-time interactions.

🛠 Technologies Used

  • Scrapy, BeautifulSoup (Web Scraping & Data Extraction)
  • Python, Hugging Face Transformers, LoRA (Fine-Tuning & Model Training)
  • Pinecone, Retrieval-Augmented Generation (RAG) (Enhanced Response Retrieval)
  • Gradio, Hugging Face Spaces (Chatbot Deployment)

📸 Screenshots


The AROL Group Smart Chatbot demonstrates the integration of LLMs, retrieval-augmented generation, and fine-tuning to create a domain-specific, AI-powered assistant. This chatbot enhances knowledge retrieval for AROL Group’s products and services while ensuring efficient and accurate responses.