
Steven (Jiaxun) Tang
Steven is an operating system researcher and an experienced software engineer.
He is currently a Ph.D. candidate at UMass Amherst.
He is enthusiastic about exploring the frontiers of software technology and building useful tools and products for users. He is passionate about his research and work, always enjoying the process of self-improvement, and loves to work in exceptional teams.
He is currently a Ph.D. candidate at UMass Amherst.
He is enthusiastic about exploring the frontiers of software technology and building useful tools and products for users. He is passionate about his research and work, always enjoying the process of self-improvement, and loves to work in exceptional teams.
Research Expertise
Machine learning system and software system performance profiling and optimization
Experience
ML Engineer Intern
May 2025
Incoming
Firefly @ Adobe
TBA.
Research Intern
Feb 2024 - Aug 2024
5 mos
US Infrastructure System Lab @ Bytedance
Diagnosing GPU memory related perforamce problems with my research work MLInsight.
Research Intern
May 2023 - Nov 2023
5 mos
US Infrastructure System Lab @ Bytedance
Diagnosing GPU memory related perforamce problems with my research work MLInsight.
RA/TA
Dec 2020 - Present
4 yr 5 mos
Mass Lab @ UMass Amherst
Research training in the MLSys performance analysis and optimization area.
Research Intern
Sep 2019 - Jan 2020
4 mos
PeLab @ SJTU
Working on a machine learning and chemical engineering interdisciplinarity research project.
RA
Sep 2016 - Feb 2019
2 yr 5 mos
IMC Lab @ SUES
Attend innovation and entrepreneurship competitions and work on computer vision commercial projects.
President of the SU
Nov 2014 - Nov 2015
1 yr 0 mos
SG High School
Student-elected president of the 46th student union. Led the digital transformation of the union. Co-chaired the student autonomous committee.
Featured Projects
MLInsight: Hierarchical Performance and Memory Analysis for ML Programs
In Progress
A lightweight, code-change-free profiler that can diagnose GPU memory-related performance problems. Joint work with ByteDance.
Scaler: Efficient and Effective Cross Flow Analysis
ASE'24
We built a light-weight profiler that achieves high data collection frequency (~700x more frequent than perf), low-performance overhead (~190x lower than bpftrace), and low memory overhead (~7x lower than perf) at the same time while being able to reveal performance problems effectively through the proposed cross-flow analysis method. Joint work with ByteDance.
CachePerf: A Unified Cache Miss Classifier via Hybrid Hardware Sampling
Sigmetrics'22
We built a profiler that correctly identifies different cache misses (false sharing/conflict miss/capacity miss) in one tool while maintaining low-performance overhead (14%).
Machine learning to assist filtered two‐fluid model development for dense gas–particle flows
AIChE'20
Trained machine learning model to improve CFD simulation model to balance accuracy and efficiency. Designed and developed a data linker to efficiently integrate the developed model into FLUENT simulation software.
High-precision cable sheath measurement instrument
Commercial Project
Designed and developed half of the computer vision algorithms to automatically perform quality inspection of electric cables. Guang Dong Nan Yang Cable Holding Co., Ltd purchased the system.
XtTech Cloud
Amateur Project
A self-developed hybrid cloud solution based on various open-source technologies. The backbone of XtTech. It provides robust and reliable data storage & backup, web hosting, virtualization, performance monitoring, and other services. This website is also a part of XtTech Cloud.
Mass Lab IT Infrastructure
Amateur Project
A cluster of services providing unified lab machine access securely to users. Provides two factor authentication-based SSH login, web-based remote desktop, network storage, uptime monitoring, and security auditing capability. Serves Mass Lab, Prof. Hui Guan, Prof. Xuan Zhang, and Prof. Sandip Kundu's research group.
ECE670 Autograder
Amateur Project
A self-implemented project autograder for the Advanced System Software Design at UMASS Amherst (ECE670). It has code judgers, a Python plugin system, email feedback & monitoring mechanisms, and an OTA update mechanism. The system served 25+ students for over three months.
XtTech Mobile Office
Amateur Project
A self-modified SUV with a monitor, a low-cost 5G/Starlink aggregated network based on mptcp, a parking air conditioner/heater, a 2kWh battery that charges in-motion, and level-2 autonomous driving. It functions as an office/studio operable in any weather condition while maintaining the appearance of a normal vehicle. Suitable for both everyday use and adventure.
Patents
63/281,942 Cacheperf patent (3rd inventor, in application)
CN109032125A An computer vision navigation method of AGV (2nd inventor)
CN109009074A Cardiac disease early warning device based on deep learning (4th inventor)
Awards
UMASS ECE department scholarship 23050 USD
2020 Shanghai excellence college graduates honor
First place (1/25 teams) in 2018 Cross-Strait College Students Innovation and Entrepreneurship Competition (1st participant)
First place (1/675 teams) in 2018 Shanghai Computer Application Ability Competition, Smart City Group (1st participant)
Silver award for 2018 “Chuang Qing Chun” Shanghai University Student Entrepreneurship Competition (2nd participant)
First place in 2018 Xipu Enterprise Scholarship
2017 Shanghai College Student Innovation and Entrepreneurship Training Program funding 9000 RMB (1st participant)
Education
Computer Engineering PhD
Sep 2021 - Present
3 yr 8 mos
COE @ UMASS Amherst
Computer Science BE
Sep 2016 - Jun 2020
3 yr 10 mos
SEEE @ SUES