AI Chatbot & Search Feature

AI Chatbot & Search Feature Final Studies Project for Kerdos Energy: • Developed an intelligent RAG-powered chatbot with semantic search capabilities to enhance user engagement with sustainable energy solutions. • Implemented in Python using Hugging Face models, leveraging retrieval-based search for the search bar and RAG architecture for chatbot responses. • Managed the project with limited resources, selecting free solutions and optimizing performance. • Tested OpenSearch before choosing Weaviate as the primary database for storing document embeddings. • Processed provided PDFs, converting them to Markdown with Docling to preserve text structure, then chunked the content for efficient retrieval. • Explored hybrid search techniques using clustering methods for retrieval, implemented reranking strategies, and experimented with chain-of-thought reasoning for improved responses. • Containerized the application using Docker and integrated Mistral LLM for French comprehension and context-specific solutions. 💡 Skills Gained: • In-depth RAG comprehension and end-to-end development capabilities • Advanced understanding of chunking and reranking strategies • Experience with chain-of-thought reasoning for LLMs • Hands-on knowledge of Weaviate, OpenSearch, Docling • Practical expertise in resource-efficient AI development, Dockerization, and semantic search implementation • Improved ability to deliver AI products to a company, addressing real-world needs and aligning solutions with stakeholder requirements