Retrieval Augmented Generation - Giving LLMs access to source documents

Continuing the theme of something new/interesting being released daily - we'll be looking at a subject that has been seeing a lot of innovation recently : Retrieval Augmented Generation ("RAG").

Talks:

"A Tour through the World of Retrieval Augmented Generation" - Sam Witteveen

In this talk Sam will go through some of the key concepts and techniques of RAG and how they can be embellished to get better results for your project, using frameworks like LangChain and Llama Index.

"RAGcipe" - Leonard Loo

Leonard will dive deep in his demo, "RAGcipe", that leverages a LLM for a personalized recipe experience. He'll demonstrate how a Self-Querying Retriever selects his daily recipe, how Ensemble Retrieval (combining Exact and Semantic Search) shortlists recipes based on specific ingredients, and how the integration of OCR and the LLM allows for the addition of handwritten recipes into the database.

"The cheapest RAG MVP?" - Peng Zhao

It’s important to launch your MVP quickly. You can do this now within 1 week, and for 0 cents. Peng will walk through step by step on how to build an RAG application (with a nice FE UX) using Google Drive, Hugging-face and Together, all by following a secret YouTube channel, along with a pain point I experienced in semantic embedding.

"What's Next in RAG" - Martin Andrews

Current RAG can be framed as an engineering approach to getting LLMs to access the right documents. But a more ideal solution would be to have the LLMs ask for the right documents - something they are not really trained for. Martin will cover some of the recent research heading in the direction of making LLMs actively seek out information for themselves.

***

Talks will start at 7:00 pm and end at around 8:50pm, at which point people normally come up to the front for a bit of a chat with each other, and the speakers.

As always, we're actively looking for more speakers - both '30 minutes long-form', and lightning talks. For the lightning talks, we welcome folks to come and talk about something cool they've done with keras_core, TensorFlow, PyTorch, JAX and/or Deep Learning for 5-10mins (so, if you have slides, then #max=10). We believe that the key ingredient for the success of a Lightning Talk is simply the cool/interesting factor. It doesn't matter whether you're an expert or an enthusiastic beginner: Given the responses we have had to previous talks, we're sure there are lots of people who would be interested to hear what you've been playing with. If you're interested in talking, please just introduce yourself to Martin or Sam at one of the events.