“Make your LLMs smarter with Open Source RAG” – Nikolay Stoitsev

This talk explores the limitations of Large Language Models (LLMs), particularly their knowledge being fixed at the time of training and restricted to the training data. The speaker discusses Retrieval-Augmented Generation (RAG) as a method to enhance LLMs by enabling the integration of real-time, domain-specific context during queries. Key topics include the inadequacy of simply prompting harder, the open-source ecosystem involving vector databases and frameworks, and practical design considerations such as chunking, embeddings, and retrieval strategies. Attendees will gain insights on setting up their own open-source RAG pipeline to enhance the intelligence of LLMs.