12-07, 17:30–18:00 (UTC), Machine Learning Track
With LLM hype growing ever greater, almost every company is racing to create their LLM application, whether it's an internal tool to boost productivity, or a chat interface for their product.
However, if your product or domain isn't fully generic, you'll probably hit a lot of challenges that make deploying your LLM application a meaningful risk.
In this talk, I'll discuss the main challenges in customizing and evaluating LLMs for specific domains and applications, and suggest a few workflows and tools to help solve for those challenges.
This talk will cover the following topics:
1. LLMs and how to use them
1. Reasons to adapt LLMs to specific use cases
1. Challenges in customizing LLMs
1. Challenges in evaluating LLMs
1. Workflows to customize LLMs – when should you take which approach
1. Workflows to evaluate LLMs – how can you reduce risks when deploying custom LLM applications
No previous knowledge expected
With a background combining Machine Learning, Software Engineering, Physics, and design – Dean applies a multi-disciplinary to building products for machine learning and AI teams.
Dean is the CEO and co-founder of DagsHub, a platform for machine learning & AI teams that lets them build better models and manage their project's data, models, experiments, and code effectively—combining popular open-source tools and formats to create a central source of truth for AI projects.
Dean is also the host of the MLOps Podcast, where he speaks with industry experts about getting ML models to production.