All Things LLM: Vector Stores, Knowledge Graphs & Models
Join the Haystack, neo4j, Qdrant and mixedbread.ai teams for an evening of talks, drinks, snacks, and lots of socializing.
📆 Agenda
6PM: Doors Open
6:30-7PM: Mingle, grab a drink, have some food
7-8:30 PM: Talks begin with an intro from your MC: Bilge Yücel
8:30-9:30: Networking and more drinks/food
9:30-10PM: vacate the venue
🎙️Talks
1. Opening Notes from Haystack - by Bilge Yücel
Some information about the meetup and Haystack by Bilge Yücel. Get to learn about Haystack 2.0 and all the cool ways you can customize your AI pipelines.
2. Superpower your GenAI-Applications with GraphRAG using Haystack and Neo4j by Michael Hunger
As you might have experienced, LLMs are powerful but not always trustworthy assistants. With a combination of a knowledge graph and vector search, you can provide the LLM with the correct, relevant context information it needs to answer your user's questions. GraphRAG is an advanced RAG (retrieval augmented generation) pattern in the Haystack-Neo4j integration.
In this talk, we'll explain and demonstrate the building blocks of such an approach and show an example of code and live in action.
Of course, nothing is perfect; it's important to walk through the challenges of building such GenAI apps and how to address them.
3. RAG Evaluation in Action: Building, Tackling Cold Start Challenges, and Optimizing Your RAG with Qdrant and RAGAS by Atita Arora
As Retrieval Augmented Generation (RAG) continues to make significant strides across diverse industries, the need to elevate its performance has become paramount. In this session, we’ll dive deep into the nuances of the process of building a Documentation RAG using Qdrant as the knowledge store, evaluating RAG pipelines and leveraging experimentation to make well-informed decisions. We'll demonstrate how the open-source evaluation framework, RAGAS, serves as a powerful ally in refining RAG solutions. But wait, there's more! We'll tackle head-on the persistent challenge of the cold start problem in building evaluation datasets. Join us as we discuss proven strategies to navigate this obstacle seamlessly, ensuring your RAG journey is successful. Experience all of this in our engaging workshop-style session with code-walkthrough and immerse yourself in practical learning.
4. Building the future of Neural Search: How to train a SOTA embeddings Model by Aamir Shakir
Neural Search plays a crucial role in Retrieval Augmented Generation (RAG) and a lot of different AI use cases. In this talk we will talk about the future of neural search, interesting challenges we are tackling and we will explain how we build our SOTA embedding model, which helps to build high quality RAG systems on scale.