Looking to dive into cutting-edge research in efficient AI, MLSys, or AI4Science? Don’t miss the chance to work with the brilliant Hongyi Hongyi Wang —an exceptional collaborator and advisor. Apply now for PhD openings in his group and shape the future of AI!

Please check our new work efficient MoE inference by mitigating the unbalanced workloads among experts.

Excited to be part of the team behind Arctic-Text2SQL-R1, now #1 on the BIRD leaderboard with +2.8% over prior SOTA! Huge thanks to our collaborators. More powerful Text-to-SQL models coming soon — stay tuned!

First fellowship from our group! 🎉 Big congrats to Guoheng Sun & Shuai He for winning the Qualcomm Innovation Fellowship! Their work looks at structural redundancy in Transformers—offering fresh ideas for efficient model design, esp. for Edge AI. Thanks Qualcomm for support!

This research was only possible thanks to a fantastic team of collaborators from Snowflake, Univ. of Maryland, and UC San Diego. Thank you to everyone involved, including Zhewei Yao, Guoheng Sun, Hao Zhang, Ang Li, and Yuxiong He.

Two UMD PH.D. Students Awarded Qualcomm Innovation Fellowship: Shwai He and Guoheng Sun Receive Prestigious Award ece.umd.edu/news/news_stor…

Knowledge makes the world so much more beautiful.