SF Unstructured Data Meetup August 5 2024

🎥 Once a month, we'll meet, socialize, and hear speakers present topics on unstructured data and generative AI. This event was sponsored by Zilliz.

Timeline:
00:05 - Introduction to Unstructured Data and the GenAI movement with Host Chris Churilo
05:35 - Speaker Robertson Taylor, A Different Angle: Retrieval Optimized Embedding Models
36:56 - Speaker Charles Xie, From Dev to Prod: Vector Database Made Easy
01:18:27 - Speaker Aamir Shakir, Building the Future of Neural Search: How to Train State-of-the-Art Embeddings

~~~~~~~~~~~~~~~ CONNECT ~~~~~~~~~~~~~~~

~~~~~~~~~~~~~~ MEETUP VIDEO CONTENTS ~~~~~~~~~~~~~~
Host: Chris Churilo

1. Speaker: Robertson Taylor
Title: A Different Angle: Retrieval Optimized Embedding Models
Abstract: When doing embedding retrieval, we rarely want just one piece of information from our document corpus. Whether we’re searching through academic papers or an ecommerce product catalog, we want multiple results which are ranked based on the query.
Marqo’s fine-tuning approach, Generalized Contrastive Learning (GCL), moves beyond the binary relationships used to train existing models. GCL introduces both rank and query-awareness to text-only and multimodal models, significantly improving relevance, especially when using user-interaction data.
In this talk, we’ll discuss how GCL works, and ways people are using it in the real world.

2. Speaker: Charles Xie
Title: From Dev to Prod: Vector Database Made Easy

3. Speaker: Aamir Shakir
Title: Building the Future of Neural Search: How to Train State-of-the-Art Embeddings
Abstract: Neural search plays a crucial role in Retrieval Augmented Generation (RAG) and various other AI use cases. In this talk, we will discuss the future of neural search, explore interesting challenges we are addressing, and explain how we build our state-of-the-art embedding model, which helps to develop high-quality RAG systems at scale.