18.10.2025 –, Зала A
Език: Български
Large Language Models (LLMs) are powerful, but they come with a big limitation: their knowledge is frozen at training time and is limited to what was in the training data. Retrieval-Augmented Generation (RAG) changes that by allowing models to pull in fresh, domain-specific context at query time.
In this talk, we’ll explore how you can build more intelligent and more useful AI systems by combining LLMs with open-source tools for RAG. We’ll cover:
- Why “just prompting harder” or having a longer context isn’t enough.
- The open-source ecosystem: from vector databases to frameworks.
- Practical design choices: chunking, embeddings, retrieval strategies, and evaluation.
At the end of the talk, you’ll have a clear understanding of how to set up your own open-source RAG pipeline and make your LLMs not just bigger, but truly smarter.
Nikolay Stoitsev leads the technical organization at the industry-redefining scale-up StorPool. His teams build and operate the blazing fast, highly available, and super scalable distributed storage platform powering critical applications worldwide. Nikolay is obsessed with building great teams and software products. He has worked at Uber, VMware, and startups across different industries. He loves to share knowledge and experience in talks and podcasts.