Jianan (Oscar) Ji

I’m currently a second-year master student at Carnegie Mellon University. I’m a member of the CMU Catalyst research group, and am fortunate to work under the esteemed guidance of Prof. Zhihao Jia.

My research interests lie at the intersection of machine learning systems and large language models. I believe that research and engineering, systems and algorithms are equally indispensable in this exciting era of LLM. Currently, my primary work involves breaking the boundaries of traditional kernels by fusing them into a “megakernel” to push hardware utilization to the limit.

Prior to CMU, I graduated with a Bachelor’s degree in Computer Science from Peking University, where I worked with Prof. Tong Yang on network system and advanced algorithms for data stream.

news

Apr 21, 2026	Excited to share that our paper MPK: A Compiler and Runtime for Mega-Kernelizing Tensor Programs has been accepted to OSDI ‘26! I’m grateful to be a co-first author of this work.
Jun 18, 2025	Our first version of Mirage Persistent Kernel is released! We developed a compiler that automatically transforms LLM inference into a single megakernel — a fused GPU kernel that performs all necessary computation and communication in one launch. This end-to-end GPU fusion approach reduces LLM inference latency by 1.2-6.7x. Have a look at the blog and the codebase.
May 31, 2025	Excited to share that our paper Mirage: A Multi-Level Superoptimizer for Tensor Programs is accepted by OSDI ‘25 !

latest posts

Jul 05, 2026	How the “Correct” Way to Align CUDA Pointers Can Slow Down Shared-Memory Loads
Sep 07, 2025	The Catastrophe of `#pragma unroll` in CUDA Programming

selected publications

OSDI 26

MPK: A Compiler and Runtime for Mega-Kernelizing Tensor Programs

Xinhao Cheng^*, Zhihao Zhang^*, Yu Zhou^*, Jianan Ji^*, Jinchen Jiang, Zepeng Zhao, Ziruo Xiao, Zihao Ye, Yingyi Huang, Ruihang Lai, Hongyi Jin, Bohan Hou, Mengdi Wu, Yixin Dong, Anthony Yip, Zihao Ye, Songting Wang, Wenqin Yang, Xupeng Miao, Tianqi Chen, and Zhihao Jia

In 20th USENIX Symposium on Operating Systems Design and Implementation (OSDI 26), Jul 2026

arXiv Bib HTML Code

@inproceedings{cheng2026mpk,
  title = {{MPK}: A Compiler and Runtime for {Mega-Kernelizing} Tensor Programs},
  author = {Cheng, Xinhao and Zhang, Zhihao and Zhou, Yu and Ji, Jianan and Jiang, Jinchen and Zhao, Zepeng and Xiao, Ziruo and Ye, Zihao and Huang, Yingyi and Lai, Ruihang and Jin, Hongyi and Hou, Bohan and Wu, Mengdi and Dong, Yixin and Yip, Anthony and Ye, Zihao and Wang, Songting and Yang, Wenqin and Miao, Xupeng and Chen, Tianqi and Jia, Zhihao},
  booktitle = {20th USENIX Symposium on Operating Systems Design and Implementation (OSDI 26)},
  year = {2026},
  month = jul,
  address = {Seattle, WA},
  pages = {1909--1926},
  isbn = {978-1-939133-55-7},
  publisher = {USENIX Association},
}

OSDI 25

Mirage: A Multi-Level Superoptimizer for Tensor Programs

Mengdi Wu, Xinhao Cheng, Shengyu Liu, Chunan Shi, Jianan Ji, Kit Ao, Praveen Velliengiri, Xupeng Miao, Oded Padon, and Zhihao Jia

Jul 2025
ICDE 25

DaVinci Sketch: A Versatile Sketch for Efficient and Comprehensive Set Measurements

Yanshu Wang^*, Jianan Ji^*, Chao-Hsuan Liu, Hengyang Zhou, and Tong Yang

In 2025 IEEE 41st International Conference on Data Engineering (ICDE), Jul 2025
SIGMOD 24

Chainedfilter: Combining membership filters by chain rule

Haoyu Li, Liuhui Wang, Qizhi Chen, Jianan Ji, Yuhan Wu, Yikai Zhao, Tong Yang, and Aditya Akella

Proceedings of the ACM on Management of Data, Jul 2023
ICDE 24

Bitmatcher: Bit-level counter adjustment for sketches

Qilong Shi, Chengjun Jia, Wenjun Li, Zaoxing Liu, Tong Yang, Jianan Ji, Gaogang Xie, Weizhe Zhang, and Minlan Yu

In 2024 IEEE 40th International Conference on Data Engineering (ICDE), Jul 2024