GenAI Cracow #24 - RAG Systems
Throwback Monday - during this edition we will back to our roots again - RAG-based systems
Agenda
Opening Ceremony
Search Small, Return Big: Building a Parent Document Retriever by Konrad
Lightning Talks - 3 slots available
Q&A
Quiz
Networking
Abstract
Main talk:
Search Small, Return Big: Building a Parent Document Retriever
RAG systems force a trade-off: small chunks embed well but lose context, while large chunks preserve meaning but reduce search accuracy. Parent Document Retriever solves this by separating what you search from what you return indexing small semantic chunks for precision, while retrieving larger parent documents for rich LLM context.
We will compare naive RAG with PDR, break down the dual-store architecture, discuss when it works (and when it fails), and build a working implementation using LangChain with persistent storage.
Right after the main presentation, we will open 3 Lightning Talks slots for the community.
A Lightning Talk is a short 5–15 minute, tech-focused session where you can share practical experience with Generative AI in a concise and informal format.
This slot is open to different types of contributors:
Project builders who want to show how their GenAI project or prototype works
Geeks using Generative AI in their daily work and willing to share their lessons learned
Problem owners who want to present a challenge and get feedback or alternative ideas from the community
Please note:
This is a technical session
No marketing or sales presentations
No product pitches
All Lightning Talk speakers are validated before the event.
It does not need to be perfect or polished, and you do not need a PowerPoint presentation.
Your Lightning Talk can be very simple, for example:
a walkthrough of a small code snippet
a quick look at a codebase
a Jupyter notebook walkthrough
...or just a short technical talk followed by discussion
The goal is sharing knowledge and practical experience, not perfect slides or formal presentations.
How to apply
Register for the event on Luma and mark your willingness to give a Lightning Talk
contact Konrad Bujak directly via LinkedIn
Speakers
Konrad Bujak
Generative AI Engineer and Product Manager, working in IT for the past 14 years. MSc. in AI Engineering. Founder @ CEE AI Hub, AI venture building and local communities booster. Sometimes also a private consultant helping companies implementing Generative AI features / products and building product (engineering) teams. IoT and homelab enthusiast.
LinkedIn | Github
Partners
Grand Parade
Grand Parade is a technology hub and part of evoke, one of the world’s leading betting and gaming companies. The Group owns and operates internationally renowned brands, including William Hill, 888, and Mr Green.
Beyond combining sports, technology and entertainment, its mission is to support the IT community.
CEE AI Hub