Advanced Retrieval Methods for RAG
When leveraging RAG, to improve retrieval is to improve augmentation. That is why there has been a proliferation of retrieval algorithms to improve overall application performance from both accuracy and cost perspectives.
It is important to retrieve the right information and order-rank that information within the prompt while maintaining a balance of being useful without being duplicative.
In this event, we will break down the retrieval algorithms that AI Engineering practitioners should know and have at hand within their toolbox. Algorithms known to provide greater precision and results at the retrieval step of RAG include the Parent Document Retriever, Self Query Retriever, Contextual Compression, and Time-Weighted Vector Store. We will move through each of these one by one:
Parent Document Retriever: small chunks are good. Big chunks are good. Do both.
Self-Query Retriever: This function constructs additional queries from the user’s initial query, breaking one question into many.
Contextual Compression: leverages document compression and retrieval to speed up the process of finding information relevant to queries that have never been seen before
Time-Weighted Vector Store: this method relies on the fact that the most often accessed data is typically the most useful
Once retrieval is complete, it is time for Augmentation. Here, reranking is often done to down-select and order-rank the top k results. To this end, we will look at the leading ranking method available today: Cohere’s Rerank. Moreover, we'll discuss other ways that we can do reranking.
Finally, a quantitative comparison will be done across methods to assess the relative performance of retrieval techniques using the RAG assessment (RAGAS) framework for a specific use case.
Finally, since there are also more general and flexible ideas about retrievers and retrieval, including methods like Multi-Vector retrieval and countless ways to incorporate agentic reasoning, we will spend some time discussing these and their utility as well! As always, we’ll cover everything you need, from concepts to code, and you’ll leave with a notebook to implement the algorithms on your next use case!
Who should attend the event:
AI Engineers who build RAG applications and optimize them for specific use cases
AI Engineering leaders interested in improving RAG prototype performance
LLM Practitioners who want to build at the open-source edge with LangChain
To prepare for this session, watch:
Speakers:
Dr. Greg Loughnane is the Co-Founder & CEO of AI Makerspace, where he is an instructor for their AI Engineering Bootcamp. Since 2021 he has built and led industry-leading Machine Learning education programs. Previously, he worked as an AI product manager, a university professor teaching AI, an AI consultant and startup advisor, and an ML researcher. He loves trail running and is based in Dayton, Ohio.
Chris Alexiuk is the Co-Founder & CTO at AI Makerspace, where he is an instructor for their AI Engineering Bootcamp. Previously, he was a Founding Machine Learning Engineer, Data Scientist, and ML curriculum developer and instructor. He’s a YouTube content creator YouTube who’s motto is “Build, build, build!” He loves Dungeons & Dragons and is based in Toronto, Canada.
Follow AI Makerspace on LinkedIn and YouTube to stay updated about workshops, new courses, and corporate training opportunities.