Dun Liang, from Changsha, Hunan, was admitted to Tsinghua University's Department of Computer Science through the Informatics Olympiad. He won an ACM-ICPC regional gold medal and championships in ISC/SC international student supercomputing competitions. During his Ph.D. under Academician Shi-Min Hu, he led the development of the Jittor deep learning framework (launched in 2018), the first framework supporting dynamic compilation, with related projects exceeding 10,000 stars on GitHub. After graduation, he founded and served as CEO of Beijing Fitten Technology, driving multiple AI Infra and AI Coding products to production. He is currently a Senior Engineer in NVIDIA DevTech, focusing on LLM training performance optimization and Megatron Core development.
Senior Engineer, AI DevTech · NVIDIA
AI Infra, LLM training optimization, Megatron Core Developer
CEO · Beijing Fitten Technology (Startup)
Led product strategy and delivery for multiple AI Infra and AI Coding products.
Postdoctoral Researcher · Tsinghua University, Department of Computer Science and Technology
Core founder and contributor of the Jittor deep learning framework.
Research Intern · MEGVII
MegBrain Developer
Tsinghua University · Ph.D. (Computer Science and Technology)
Advisor: Academician Shi-Min Hu
Tsinghua University · B.Eng. (Computer Science and Technology)
Changjun High School
Champion (Rank 1), SC16 Student Cluster Competition
Champion (Rank 1), ISC15 International Supercomputing Competition
Technology Innovation Award, Tsinghua CST
ACM-ICPC Hangzhou Regional Gold Medal
10th in NOI Team Selection, 43rd in National Finals
We fought side by side in Tsinghua's supercomputing team and won the SC15 Student Cluster Championship, with Prof. Jidong Zhai as our team coach. This story was later featured in Tsinghua University's 2021 admissions film, "Breaking Through Walls" (Link).
On Jan 26, 2026, I took a photo with Jensen Huang at the NVIDIA annual meeting. Coincidentally, it was my 32nd birthday, and I received a memorable birthday wish.
In June 2021, I defended my PhD thesis. I am deeply grateful for the guidance of Academician Shi-Min Hu and the intense years spent building the Jittor framework.
In early summer 2021, I took this photo in Tsinghua with my wife, a PhD in biology from Peking University—a shared memory across two great campuses.
A 2025 photo with my startup teammates. We have been through storms together, and that journey made us value companionship even more.
Graduation photo of CST25 in 2016. Youth fades, friendship stays.
At Tsinghua's anniversary celebration, the Wushu Association and Wushu Team performed together under the guidance of Prof. Yulin Wang.
I revisited the Changjun informatics group and met my high-school Olympiad coach, Mr. Xiang Qizhong. Looking back, I am most grateful to the people who lit the spark at the beginning.
Traffic-sign detection and classification in the wild
Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
Citations 1375 · 2016
Jittor: a novel deep learning framework with meta-operators and unified graph execution
Science China Information Sciences 63 (12), 222103
Citations 175 · 2020
Are we ready for a new paradigm shift? a survey on visual deep mlp
Patterns 3 (7)
Citations 129 · 2022
Recursive-nerf: An efficient and dynamically growing nerf
IEEE Transactions on Visualization and Computer Graphics 29 (12), 5124-5136
Citations 72 · 2022
Efficient and flexible deformation representation for data-driven surface modeling
ACM Transactions on Graphics (TOG) 35 (5), 1-17
Citations 72 · 2016
Lane detection: A survey with new results
Journal of Computer Science and Technology 35 (3), 493-505
Citations 71 · 2020
LineNet: A zoomable CNN for crowdsourced high definition maps modeling in urban environments
arXiv preprint arXiv:1807.05696
Citations 24 · 2018
Can attention enable MLPs to catch up with CNNs?
Computational Visual Media 7 (3), 283-288
Citations 21 · 2021
Jrender: An efficient differentiable rendering library based on Jittor
Graphical Models 130, 101202
Citations 1 · 2023