I can download more RAG? It's Pydata Copenhagen!

Ever wanted to download more RAG? Now's your chance, as the theme of the day is RAG and more RAG, with two great talks on improving your RAG game. Join us in this (probably) world-first joint-venture between Pydata x GDG x Google AI!

Sign up here!
https://rsvp.withgoogle.com/events/py-data-x-gdg-x-google/home

NB! You must sign up at the signup link above in order to join this event - only signing up on Meetup.com will not get you in the door!

Agenda:
18:00 Arrive, hang out, get some snacks, beer, soda,
18:30 Finetuning Embedding Models Locally: Building an AI Data product without having to ask your IT Department
Dilovan Celik, Data Engineer and ML Hobbyist

LLMs show great potential to generate value across many industries. However open source algorithms might be too general, and training your own model can be very resource-intensive. This talk will be a guide on how to fine-tune an embedding model for building a RAG setup. It will cover several techniques, making building your own AI powered data products less resource intensive.

19:10 They level up RAG systems with this one weird trick. VectorDB companies hate them!
Jakob Pörschmann, AI & ML Customer Engineer - Google AI

Simple RAG works well for some cases, but quickly fails if queries turn to be more complex. Enter Graph RAG, a powerful extension that leverages knowledge graphs the real world relationships of the underlying data to produce more meaningful answers. We'll explore the key steps of a Graph RAG pipeline, including graph extraction, community detection, and community report generation, and discuss how these techniques empower LLMs to answer "global" questions that traditional RAG struggles with. We'll also examine practical considerations, such as storage options and optimization strategies for efficiency and scalability.

19:45:00 More socializing, discussions, hanging out
21:00 Doors close

Sponsors:
Thanks to NumFocus for running Pydata!

Special thanks to Google for hosting us!