BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Training LLMs Anywhere: Enabling Large-Scale Decentralized Learnin
 g on Your Mobiles Devices - Dongqi Cai - Beijing University of Posts and T
 elecommunications
DTSTART:20250128T150000Z
DTEND:20250128T160000Z
UID:TALK226192@talks.cam.ac.uk
CONTACT:Sally Matthews
DESCRIPTION:Training large language models (LLMs) is often seen as a resou
 rce-intensive task\, requiring massive computational power centralized in 
 data centers. But what if you could train powerful models directly on your
  everyday devices? In this talk\, we introduce cutting-edge techniques tha
 t bring efficient LLM training to mobile and edge devices\, overcoming con
 straints like limited memory\, processing power\, and network bandwidth. W
 e present novel methods\, including adaptive federated learning and backpr
 opagation-free optimization for cross-device collaboration. These innovati
 ons empower large-scale decentralized learning\, reducing system costs whi
 le maintaining high performance and privacy. Join this talk to explore how
  this research is reshaping on-device AI\, making LLM fine-tuning practica
 l\, efficient\, and closer than ever to your fingertips.\n\nBio\nMr. Dongq
 i Cai is a fourth-year PhD student at Beijing University of Posts and Tele
 communications\, currently a visiting PhD student in Prof. Nicholas D. Lan
 e’s group at the University of Cambridge. His research focuses on effici
 ent on-device machine learning systems. He has authored 12 papers as the f
 irst or corresponding author\, including 7 in top-tier venues such as ACM 
 MobiCom\, USENIX ATC\, NeurIPS\, ACM Computing Surveys\, and IEEE Transact
 ions on Big Data. He has received multiple National PhD Scholarships and s
 erves as PC members for leading conferences's AE committee like ACM MobiCo
 m and ACM MobiSys\, while also reviewing for prestigious journals includin
 g IEEE TMC\, IEEE TSC\, and IEEE TKDE.
LOCATION:Computer Lab\, FW26
END:VEVENT
END:VCALENDAR