Unstructured Data in LLMs - South Bay Edition

This is an in-person event! Registration is required in order to get in.
​Topic: Structured Output from Unstructured Input
​
What we’ll do:
Have some food and refreshments. Hear three exciting talks about unstructured data and generative AI.

5:30 - 6:00 - Welcome/Networking/Registration
6:05 - 6:35 - Constrained Sampling from Large Language Models: Producing Structured Output, Stefan Webb, Zilliz
6:35 - 7:05 - Infinite world generation with Outlines, Cameron Pfiffer, Developer Relations Engineer, .txt
7:05 - 7:35 - BAML: Cheaper, faster, and more accurate function calling - without any fine-tuning, Vaibhav Gupta, BoundaryML
7:40 - 8:30 - Networking
​
Tech talk 1: Constrained Sampling from Large Language Models: Producing Structured Output
Speaker: Stefan Webb, DevRel, Zilliz
Abstract: Large Language Models (LLMs) are typically used to generate free-form text. However, many applications require output with specific structures or formats. This presentation introduces the concept of constrained sampling from LLMs, a method to produce output that adheres to predefined formats or structures. The key points we will cover:

• Traditional LLM usage vs. constrained sampling
• Examples of interesting applications like classification/regression and producing JSON
• Overview of how it works

We set the stage for the following two speakers who will discuss specific tools for implementing this technique: Outlines and BAML Tech

Talk 2: Infinite world generation with Outlines
Speaker: Cameron Pfiffer, .txt
Abstract: Cameron presents "LoreMaster 6000," an internally consistent world generator that infinitely builds a fictional world's history. We use Outlines to make the system programmable, and Milvus to handle lore storage and semantic search.

Tech Talk 3: BAML: Cheaper, faster, and more accurate function calling - without any fine-tuning
Speaker: Vaibhav Gupta, CEO, BoundaryML
Abstract: We'll share techniques for building agents that require function calling and how BAML, our new programming language, helps with that while providing a great developer experience.

​Who Should attend:
Anyone interested in talking and learning about Unstructured Data and Generative AI Apps.
​
When:
October 15, 2024
5:30PM

Where:
This is an in-person event. Registration is required to get into the event. Registration in advance will close 2 days before the event.
​This event is co-sponsored by Zilliz (maintainers of Milvus)