Keren Zhou

Keren Zhou

Assistant Professor

George Mason University

Biography

I am an Assistant Professor in the Department of Computer Science at George Mason University and a part-time Member of Technical Staff at OpenAI. Before joining GMU, I was a full-time Member of Technical Staff at OpenAI. I obtained my Ph.D. degree from Rice University, advised by Professor John Mellor-Crummey. Previously, I studied at Institute of Computing Technology, Chinese Academy of Sciences in Professor Guangming Tan’s PAA group. Prior to that, I was an undergraduate student at Yunnan University, advised by Professor Wei Zhou.

Interests
  • High Performance Computing
  • Parallel Algorithms
  • Program Analysis
  • Machine Learning
Education

  • PhD in Computer Science, 2022

    Rice University

  • MS in Computer Science, 2017

    Institute of Computing Technology, Chinese Academy of Sciences

  • BE in Network Engineering, 2014

    School of Software, Yunnan University

Recent News

More news…

Experience

 
 
 
 
 
OpenAI
Member of Technical Staff
OpenAI
November 2023 – Present Fairfax
Triton Compiler
 
 
 
 
 
George Mason University
Assistant Professor
George Mason University
August 2023 – Present Fairfax
Computer Architecture/Compiler/Machine Learning Systems
 
 
 
 
 
OpenAI
Member of Technical Staff
OpenAI
June 2022 – August 2023 San Francisco
Performance Optimization of Deep Learning Workloads
 
 
 
 
 
NVIDIA
Software Engineering Intern
NVIDIA
May 2021 – August 2021 Remote
Performance Profiling for Deep Learning Frameworks
 
 
 
 
 
Google
Software Engineering Intern
Google
May 2020 – August 2020 Remote
Performance Regression Analysis of Feedback-direct Optimization (FDO) Based Programs
 
 
 
 
 
Meta
Research Intern
Meta
June 2018 – August 2018 Menlo Park
Neural Network Optimization on Mobiles
 
 
 
 
 
NVIDIA
Research Intern
NVIDIA
April 2017 – July 2017 Beijing
Neural Network Quantization
 
 
 
 
 
Baidu
Software Engineering Intern
Baidu
October 2014 – February 2015 Beijing
Hadoop Workflow Optimization

Projects

*
GPA
GPA is a performance advisor for NVIDIA GPUs that suggests potential code optimization opportunities at a hierarchy of levels, including individual lines, loops, and functions. GPA uses data flow analysis to approximately attribute measured instruction stalls to their root causes and uses information about a program’s structure and the GPU to match inefficiency patterns with suggestions for optimization. GPA estimates each optimization’s speedup based on a PC sampling-based performance model.
Code
GPA
HPCToolkit
Our tool provides a profile view and a trace view for GPU-accelerated applications. The profile view identifies where GPU APIs are invoked in CPU calling context, approximates calling context for GPU execution, and analyzes instruction mix for GPU kernels. The tool traces CPU and GPU activities for a large number of processes and threads with minimal overhead.
Code DOC
HPCToolkit
Triton
Triton is a language and compiler for writing highly efficient custom Deep-Learning primitives. The aim of Triton is to provide an open-source environment for expressing tensor math workloads that offers high flexibility, developer productivity and end to end performance.
Code DOC
Triton
GVProf
We implemented GVProf, the first value profiler that locates value redundancy problems in applications running on GPU-based clusters. Our experiments show that GVProf incurs acceptable overhead and scales to large executions. GVProf provides useful insights to guide performance optimization. Under the guidance of GVProf, we optimized several HPC and machine learning workloads, obtaining speedups up to 1.93x.
Code DOC
GVProf

Recent Publications

Quickly discover relevant content by filtering publications.
Triton-Viz: Visualizing GPU Programming in AI Courses
Tejas Ramesh, Alexander Rush, Xu Liu, Binqian Yin, Keren Zhou, Shuyin Jiao
2025 Proceedings of the 56th ACM Technical Symposium on Computer Science Education (SIGCSE).

Cite Project URL

SS1: Accelerating Inference with Fast and Expressive Sketch Structured Transform
Aditya Desai, Kimia Saedi, Apoorv Walia, Jihyeong Lee, Keren Zhou, Anshumali Shrivastava
2024 The Thirty-eighth Annual Conference on Neural Information Processing Systems (NeuIPS).

Cite Project

Centimani: Enabling Fast AI Accelerator Selection for DNN Training with a Novel Performance Predictor
Zhen Xie, Murali Emani, Xiaodong Yu, Dingwen Tao, Xin He, Pengfei Su, Keren Zhou, Venkatram Vishwanath
2024 USENIX Annual Technical Conference (USENIX ATC).

Cite Project URL

FASTEN: Fast GPU-accelerated Segmented Matrix Multiplication for Heterogenous Graph Neural Networks
Keren Zhou, Karthik Ganapathi Subramanian, Po-Hsun Lin, Matthias Fey, Binqian Yin, Jiajia Li
2024 Proceedings of the 38th ACM International Conference on Supercomputing (ICS).

Cite Project DOI URL

PyTorch 2: Faster Machine Learning Through Dynamic Python Bytecode Transformation and Graph Compilation
Jason Ansel, Edward Yang, Horace He, Natalia Gimelshein, Animesh Jain, Michael Voznesensky, Bin Bao, Peter Bell, David Berard, Evgeni Burovski, Geeta Chauhan, Anjali Chourdia, Will Constable, Alban Desmaison, Zachary DeVito, Elias Ellison, Will Feng, Jiong Gong, Michael Gschwind, Brian Hirsh, Sherlock Huang, Kshiteej Kalambarkar, Laurent Kirsch, Michael Lazos, Mario Lezcano, Yanbo Liang, Jason Liang, Yinghai Lu, C. K. Luk, Bert Maher, Yunjie Pan, Christian Puhrsch, Matthias Reso, Mark Saroufim, Marcos Yukio Siraichi, Helen Suk, Shunting Zhang, Michael Suo, Phil Tillet, Xu Zhao, Eikan Wang, Keren Zhou, Richard Zou, Xiaodong Wang, Ajit Mathews, William Wen, Gregory Chanan, Peng Wu, Soumith Chintala
2024 Proceedings of the 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS).

Cite Project DOI URL

See all publications

Recent & Upcoming Talks

Proton: Adaptive and Lightweight Profiling for Deep Learning Workloads
A gentle introduction to the Proton profiler and its updates.
Mar 1, 2025 9:34 PM — 9:34 PM Las Vegas, NV
Project Slides
The Proton Dialect: An MLIR Dialect For AI Compiler GPU Kernel Profiling
Presented our work on multi-level instrumentation-based profiling for Triton kernels.
Mar 1, 2025 9:34 PM — 9:34 PM Las Vegas, NV
Project Slides
Profiling and Debugging GPU-accelerated AI Applications
Presented our research on debugging and profiling of GPU-accelerated AI applications.
Oct 24, 2024 9:41 PM — 9:41 PM Virtual
Project Slides
Proton: Introduction and Development
Presented the ongoing work on Proton
Oct 21, 2024 9:41 PM — 9:41 PM Virtual
Project Slides
Dev Tools: Proton/Interpreter
Presented the Proton and Interpreter tools in the Triton project.
Sep 17, 2024 9:41 PM — 9:41 PM Virtual
Project Slides Video
See all events

Students

  • PhD Students
    • Tejas Ramesh, GMU, 2023-
    • Bowen Cui, GMU, 2024-
    • Jihyeong Lee, GMU, 2024-
    • Hao Wu, GMU, 2024-
    • Junyu Yin, GMU, 2024-
    • Atul Khatri, GMU, 2023-2024
      • co-advised with Prof. Songqing Chen
    • Mao Lin, UC Merced, 2022-2023
      • co-advised with Prof. Pengfei Su
  • Master Students

Popular Topics

Accelerators AMD Calling Context Compiler Deep Learning Feedback-directed Optimization GPU HPC Instruction Sampling Instrumentation Interpreter Performance Performance Tool Profiler Profiling Programming Language Proton Triton Value Analysis Wait-free