Blog/Yuxiong He

Yuxiong He

Sr. Director, Software Engineering

Yuxiong He is a Sr. Director, Software Engineering, spearheading the development and research of Large Language Models (LLMs). As a pivotal co-leader of the Arctic project, she collaborates with a team of exceptional AI professionals to develop the Snowflake suite of foundational models. Her dedication to innovation is matched by her commitment to open source and open research, striving to build transformative and high-performing AI technologies. Previously, Yuxiong held the position of Partner Research and Product Manager at Microsoft, where she co-founded and led the DeepSpeed project. This industry-leading, open-source deep learning optimization library introduced groundbreaking innovations like ZeRO, 3D parallelism, and ZeroQuant. These advancements have significantly accelerated and democratized the training and inference processes of cutting-edge LLMs, making them more accessible to everyone in need. Yuxiong has published over 100 papers in major computer science conferences and journals. Her work has been recognized among the best papers at esteemed venues such as SIGIR, ICDE, WSDM, and Middleware, and her research continues to be widely applied in diverse systems and products.

MAY 07, 2026Gen AI

Building Reliable Data Science Agents with DARE-Bench and PRISM-DS

Fan Shu +6

MAR 04, 2026Gen AI

Fast and More Accurate Causal Parallel Decoding Using Jacobi Forcing

Lanxiang Hu +4

FEB 13, 2026Gen AI

Agent World Model (AWM): Infinity Synthetic Environments for Agentic Reinforcement Learning

Zhaoyang Wang +6

NOV 04, 2025Gen AI

Smarter, Faster and Snowflake-Native: Real-Time Text2SQL Behind Snowflake Intelligence

Lukasz Borchmann +5

JUN 24, 2025

Arctic Long Sequence Training (ALST): Scalable And Efficient Training For Multi-Million Token Sequences

Stas Bekman +8

JUN 03, 2025Gen AI

Inside Snowflake Intelligence: Five Pillars of Enterprise-Grade Agentic AI

Yuxiong He +7

MAY 29, 2025Gen AI

Smaller Models, Smarter SQL: Arctic-Text2SQL-R1 Tops BIRD and Wins Broadly

Zhewei Yao +2

MAY 29, 2025Gen AI

Arctic Inference with Shift Parallelism: The Fastest Open Source Inference System for Enterprise AI

Samyam Rajbhandari +8

MAY 29, 2025Gen AI

Scaling vLLM for Embeddings: 16x Throughput and Cost Reduction

Charles Xu +6

2

Yuxiong He

Building Reliable Data Science Agents with DARE-Bench and PRISM-DS

Fast and More Accurate Causal Parallel Decoding Using Jacobi Forcing

Agent World Model (AWM): Infinity Synthetic Environments for Agentic Reinforcement Learning

Smarter, Faster and Snowflake-Native: Real-Time Text2SQL Behind Snowflake Intelligence

Arctic Long Sequence Training (ALST): Scalable And Efficient Training For Multi-Million Token Sequences

Inside Snowflake Intelligence: Five Pillars of Enterprise-Grade Agentic AI

Smaller Models, Smarter SQL: Arctic-Text2SQL-R1 Tops BIRD and Wins Broadly

Arctic Inference with Shift Parallelism: The Fastest Open Source Inference System for Enterprise AI

Scaling vLLM for Embeddings: 16x Throughput and Cost Reduction

Where Data Does More