RAG Optimization Workshop
Advanced RAG strategies: Optimizing your Semantic Retrieval
Workshop #1 (Qdrant)
LLM outputs are only as good as the documents we provide for answer generation. We’ll start with an existing RAG pipeline using Qdrant and go through advanced strategies for optimizing it. Optimization constraints depend on your speed, memory, and quality requirements - pushing semantic search to the limits and combining it with different retrieval methods.
Who it is for
Developers planning to build or already building Retrieval Augmented Generation applications or anyone who wants to optimize retrieval in GenAI.
Instructor: Kacper Lukawski is a software developer and data scientist at heart, with an inclination to teach others. Public speaker, working in DevRel.
Format: Hands-on and code walkthrough
What you will learn:
Basics of vector search. The challenges of vector search based on neural embeddings.
Tweaking semantic retrieval. What to do when you need to increase search quality, reduce memory requirements, or improve speed.
Building hybrid search. Mixing different retrieval strategies to handle scenarios in which vector search fails.
Prerequisites:
Some experience in implementing RAG, with or without high-level frameworks such as Langchain or LlamaIndex
Basic familiarity with vector search and information retrieval concepts
Resources: Slides and Jupyter notebooks
Workshop #2 (Haystack)
What if you had an AI assistant that knows not only human language but also has access to REST APIs and other structured data sources?
Through practical exercises, we'll dynamically generate function calls from OpenAPI specifications, enabling LLMs to interpret queries and retrieve structured data from diverse enterprise services.
Instructor: Julian Risch is a Senior Machine Learning Engineer at deepset, where he maintains and develops the open-source LLM framework Haystack.
Format: Hands-on and code walkthrough
Who it is for: Anyone eager to explore practical applications of LLM-driven automation and unafraid to delve into Python code.
What you will learn
Building RAG pipelines with Haystack and a QdrantDocumentStore. Integration of OpenAPI-defined services with LLMs. Dynamic Function Call Generation. Using LLMs to compare Github branches and summarize code changes as an example application.
Prerequisites: Basic Python language skills. First experience with Haystack 2.0 is not required but a plus: https://haystack.deepset.ai/
Resources: Slides and Jupyter notebooks
In partnership with Haystack by Deepset.
Agenda:
16:00-17:15 —> Workshop #1 (Qdrant)
17:15-17:45 —> Break: pizza, drinks, socializing
17:45-19:00 —> Workshop #2 (Haystack)