Kai Yi (@williamyi96) profile picture
Sunnyvale, CA
Joined July 2016
Photo by Kai Yi (@williamyi96)
Unlock your potential at KAUST, ranked #1 in the Arab world by Times Higher Education.​ Work with world-class experts, access top facilities, and lead research with global impact.​ Apply now for MS and PhD programs.

1
17
1
39
11,266
🚀 Career Update Excited to start my journey at @Meta, California as a Research Scientist, focusing on model compression and inference acceleration. It’s amazing to meet so many passionate people, and I’m proud to join a team working on impactful research.
3
1
8
2,149
(3/3)A new chapter begins — grateful for the past and excited about what lies ahead! ✨
1
176
Replying to @williamyi96 @Meta
(2/3)📑 Fun fact: my U.S. work visa journey lasted 259 days — a truly challenging process. Huge thanks to my HRBP, attorneys, and everyone involved for the tremendous support to finally make it happen.
1
224
Replying to @williamyi96 @Meta
(1/3)🎓 I also recently graduated with a PhD in Computer Science from KAUST. Deep thanks to my advisor @peter_richtarik and the incredible team for their constant support over the past years — couldn’t have made it without you. 🙏
1
297
I am not attending @NeurIPSConf this year but many of my students are! Do reach out to Omar Shaikh Omar, Hanmin Li, Kai Yi (@williamyi96), Abdurakhmon Sadiev (@A_Sadiev), Kaja Gruntkowska, Egor Shulgin (@egor_shulg), Artavazd Maranjyan and Marta Pozzi! Here are our papers: check them out! Want to join my lab as an intern, MS/PhD student, PhD student or a postdoc? Apply here: apply.interfolio.com/105097 Plus read this, too: richtarik.org/i_join.html
7
35
3,168
Thanks @bozavlado for such an enthusiastic endorsement!
Me a week ago: I can easily beat GPTQ (method from 2022), and maybe after a month, I can make ~SOTA quantization. Me yesterday: F***, this work moved SOTA very far, no chance to catch it. Anyway, LLM quantization is becoming unbelievably good. arxiv.org/pdf/2405.14852
3
5
1,904
I am looking for exceptional postdocs to join my Optimization and Machine Learning Lab at KAUST (richtarik.org/i_team.html)! Apply for the #KAUST Global Fellowship! In addition, upload your files here: apply.interfolio.com/105097 @KAUST_News @AI_KAUST @KaustResearch
Join the #KAUST Global Fellowship Program, an incredible opportunity for exceptional postdocs with groundbreaking ideas to advance Saudi's research priorities. The program offers access to cutting-edge research tools, top-notch facilities and expert mentorship.

8
1
33
7,740
Happy to release PV-Tuning, a new fine-tuning technique for highly-compressed LLMs, which sets SOTA PTQ accuracy for 1-2.5bit LLMs. Joint work with @peter_richtarik's lab. arXiv: arxiv.org/abs/2405.14852 Github: bit.ly/3R2VYHl HF Models: bit.ly/4aIEgQF
13
1
41
7,344
Our paper HGR-Net has been accepted by ECCV 2022 @eccvconf! Congratulations to my collaborators, Xiaoqian Shen, Yunhao Gou, and @moElhoseiny! Your support and work really help! arXiv link: arxiv.org/abs/2203.01386 code link: github.com/WilliamYi96/HGR-N…
4
2
1
33
Meet #KAUST prospective student: Kai Yi @williamyi96 Read the full story cemse.kaust.edu.sa/news/meet… @KAUST_News @cemseKAUST
2
3
“Android TensorFlow Machine Learning Example” by @amitiitbhu blog.mindorks.com/android-te…
2
I believe the title will raise a storm of conflicts as it's so ** that just like a little gimmicky. nitter.cf/karpathy/status/876628…
"One Model To Learn Them All" another step in Google's attempt to turn all of itself into one big neural network arxiv.org/abs/1706.05137
I‘m addicted to ''Research on Machine Learning and Asymmetric Information Game''. Promote to this field with my own efforts. Fighting! – at San Francisco, CA
1
Learning Zone and Performance Zone. More time to exploring, asking, listening, and reflecting. ted.com/talks/eduardo_bricen… via @TEDTalks
3
Survey paper on Geometry of Optimization & Implicit Regularization in Deep Learning with Neyshabur, Tomioka, Srebro arxiv.org/abs/1705.03071 – at Toronto, Ontario
1
55
2
124
Maybe this idea can be safely transformed into deep reinforcement learning? Just see. nitter.cf/seb_ruder/status/86923…
New blog post on Multi-Task Learning in Deep Neural Networks sebastianruder.com/multi-tas…
1
1
I believe it will raise a storm of research on deep reinforcement learning after the winning of alphago. cheer for the road to true ai.
1
Reinforcement learning maybe a booming field which will lead the future. From now on, I will try all out to go deeper about it. Move on!💪 – at San Mateo, CA
1