DEMO: How To Align LLMs for Enterprise Applications in Snorkel Flow
Large language models (LLMs) don't work "out of the box" for enterprise applications. Data scientists must align them with training data, but that process can be expensive and laborious.
Snorkel AI product lead Marty Moesta demonstrates how the Snorkel Flow AI data development platform makes this process easier and faster than ever before.
This video is an excerpt from a longer webinar. See the full webinar here: • How to Fine-Tune LLMs to Perform Specializ...
See more Snorkel Flow product demonstrations here: • Snorkel Flow Demos: See How it Works!
Timestamps:
00:00 Introduction
00:32 Overview of LLM Systems
01:30 Development Approach
02:15 Focus on Generation Errors
02:39 Approaches to Solve Generation Errors
03:36 Aligning LLM Programmatically
04:41 Demo Introduction: Acme Credit Services
05:20 Pre-Production Environment Testing
07:14 Manual Annotation in Snorkel Flow
09:29 Importance of Collaboration
10:30 Building a Quality Model
12:48 Creating a Curated Dataset
15:30 Evaluating Model Performance
18:36 Summary of Progress
19:22 Ensuring Robustness Across Data Slices
21:30 Writing Data Slices
23:33 Redeploying Fine-Tuned LLM
24:20 Conclusion and Next Steps
#enterpriseai #largelanguagemodels #alignment
![El Cascabel [Son Jarocho] | The Mesoamerican Orchestra](https://images.mixrolikus.cc/video/O0Ln-DnBMGk)
