ICSA Colloquium Talk - 02/05/2024

Title: Status and Challenges of ByteDance’s Large Foundation Models

Abstract: Many teams and customers with ByteDance are operating on AI infrastructures for training and deploying large foundation models (including large language models and multi-modal models). This presentation will outline the system architecture and discuss recent advancements in scaling our AI infrastructure. At the end of this talk, I will explore the system and model challenges we encounter in production.

Bio: Wei Xu is the Director of the Foundational Model Lab at ByteDance, overseeing operations in both California and Seattle. Prior to ByteDance, he held the position of Engineering Director at VMWare, FutureWei, and Cisco.