|
|
Harry Xu Professor
|
|
About Me
I am a Professor in the Computer Science Department of University of California, Los Angeles. I co-founded BreezeML, a startup company that builds a software solution for GenAI risk management. In 2017, I worked at Microsoft Research as a Visiting Researcher where I created and led the development of a project that built an optimizing compiler (Niijima, SOSP'19) for Microsoft's Scope/Cosmos, a multilingual data analytics pipeline. I worked at IBM T. J. Waston Research Center as a Co-op/intern from 2008 to 2011 where I led the development of a series of runtime bloat detection/performance optimization tools for IBM's enterprise Java customers (PLDI'09, PLDI'10-a, PLDI'10-b). A recent project VQPy developed by our group has been integrated into Cisco's DeepVision platform for expressing and answering complicated video queries.
My research is in the general area of computer systems, ranging from programming languages and compilers, to runtime/operating/distributed systems and computer architecture. My current focus is on two areas: (1) infrastructures for future cloud computing (in particular, user-defined cloud) and (2) infrastructures that provide cheap scalability for ML applications (i.e., affordable AI).
I was one of the first academic researchers working on the problem of ``software bloat'', which eventually led to ONR's TPCP program. I also spearheaded the direction of optimizing large-scale data analytics with PL techniques. Our Yak GC (OSDI'16) was reimplemented and used by Huawei to power their telecommunication applications. Due to these contributions, I was awarded the 2018 Dahl-Nygaard Junior Prize, which is awarded each year to a junior researcher that has made promising contributions to the field of object orientation.
Professional
Current Projects (Some Recent Papers)
o
VQPy: Object-Orientation for Modern Video Analytics (MLSys'24)
o
Systems for Affordable Machine Learning (SIGCOMM'20, OSDI'21, NSDI'23-b, NSDI'23-c, NSDI'24-b)
o
Systems for Resource-Disaggregated
Datacenters (OSDI'20, PLDI'22-a, OSDI'22-best-paper, NSDI'23-a, NSDI'23-d, NSDI'24-a, OSDI'24-a, OSDI'24
-b)
o
System Design for Emerging (Datacenter) Hardware (PLDI'19, PLDI'20, ASPLOS'21, PLDI'21-a, ASPLOS'22-a, ASPLOS'22-b, PLDI'22-b, OSDI'22-b)
o
IoT Usablity and Security (SEC'18, FSE'20-a)
o
Graph Systems for Static Code
Analysis and Other Workloads (ASPLOS'17-a, ASPLOS'17-b, OSDI'18, EuroSys'19, PLDI'21-b, TOCS'21, EuroSys'23)
o
Big Data Systems and their
Runtime Support (SOSP'15, ASPLOS'15, OSDI'16, ASPLOS'18, SOSP'19-a, SOSP'19-b)
o
Software Bloat Detection and
Optimization (ECOOP'15, PLDI'18, FSE'20-b)
Research Group
o Jiyuan Wang (Ph.D. student starting
Fall 2019, Tsinghua Xuetang Talents Scholar, co-advised with Miryung Kim)
o Shi Liu
(Ph.D. student starting Fall 2020, Guo Moruo Scholar at USTC)
o Shan Yu
(Ph.D. student starting Fall 2022, Excellent Graduate of ZheJiang Univ., Intel DRA Award)
o Zhenting Zhu
(Ph.D. student starting Fall 2023, Yao Award, Andrew Yao Class at Tsinghua, Gold Medalist IOI 2018)
o Shu Anzai
(Ph.D. student starting Fall 2023, Funai Overseas Scholarship, University of Tokyo)
o Yicheng Liu
(Ph.D. student starting Fall 2024, Co-advised with Sam Kumar, Top Graduate of Shanghai Jiaotong U.)
Alumni
o Haoran Ma (Ph.D. 2024, first employment: Software Engineer at ByteDance AI Infra)
o Yifan Qiao (Ph.D. 2024, first employment: Postdoc Researcher, Sky Computing Lab, UC Berkeley)
o
Christian
Navasca (Ph.D. 2023, first employement: Senior ML Engineer at BreezeML)
o Pengzhan Zhao
(M.S. 2023, first employment: Co-founder and Director of Engineering at BreezeML)
o
Chenxi Wang (PostDoc 2018-2022, first employment: Associate Professor, Institute of Computing Technology, Chinese Academy of Sciences)
o
Jonathan Eyolfson (PostDoc 2018-2022, first employment: Assistant Professor in teaching stream, University of Toronto)
o
John
Thorpe (Ph.D. 2022, first employment: Co-founder and VP of Product at BreezeML Inc.)
o
Arthi Padmanabhan (Ph.D. 2022, first employment: Assistant Professor, Harvey Mudd College)
o Usama Hameed (M.S. 2022, first employment: Apple)
o
Yuanqi
Li (dropped out of Ph.D. due to startup)
o
Khanh Nguyen (Ph.D. 2019, first employment: Google & Assistant Professor @ TAMU, Google Ph.D. Fellowship, Facebook Ph.D. Fellowship
Finalist)
o
Kai Wang (Ph.D. 2019, first employment: Research Scientist @ Facebook System&Infrastructure)
o
Zhiqiang Zuo (PostDoc
2015--2018; Assistant Professor @ Nanjing University)
o
Lu Fang (Ph.D. 2017, first
employment: Research Scientist @ Facebook AI)
o
Keval Vora (Ph.D. 2017, Co-advised
with Rajiv Gupta, first employment: Assistant Professor @ Simon Fraser
University)
o
Shen Teng (M.S. 2021, first employment: Amazon)
o
Gaohong Liu (M.S. 2021, first employment: RED (Xiaohongshu E-commerce))
o
Cheng
Cai (M.S. 2020, first employment: Facebook)
o
Bojun
Wang (M.S. 2019, first employment: Arista Networks)
o
Guanzhou Hu (CSST 2019, currently Ph.D. student at UWisc)
o
Tim
Nguyen (B.S. 2017, first employment: Facebook)
o
Aftab
Hussain (M.S. Fall 2017)
o
Matthew
Hartz (B.S. 2015)
o
Jianfei Hu
(M.S. 2015, first employment: Google)
o
Chengxi
Li (2017 summer intern from Beckman
High, college: Berkeley CS)
o
Emily
Hu (2016 summer intern from Northwood
High, college: Berkeley CS)
o
Lawrence
Xu (2016 summer intern from University High,
college: Purdue CS)
o
Louis
Zhang (2014 summer intern from Troy High,
college: Berkeley CS)
o
Ankur
Gupta (2014 summer intern from University
High, college: Regent's Scholar at UCSD CS)
o
Wendy
Wei (2013 summer intern from University High,
college: MIT CS)
o
Allen
Min (2012 summer intern from Whitney High, college: UCI CS)
o Jonathon Tsai (2012 summer
intern from Whitney High,
college: UCI CS)
Teaching
o
CS 239: Large-Scale Machine Learning Systems (Spring 2020)
o
CS 111: Operating Systems (Spring 2019, Winter 2020, Winter 2021)
o
CS 214: Big Data Systems (Fall 2018, Fall 2019, Fall 2020)
Recent Services
Released Software Systems (Our Github Repos)
- Atlas: A hybrid data plane for remote memory applications
- DRust: An ownership-based DSM system
- VQPy: An object-oriented approach to video analytics queries
- Midas: Soft Memory as an OS service
- Hermit: A low-overhead and high-throughput swap system for remote memory via feedback-directed asynchrony
- Canvas: An isolatd and adaptive swap system for remote memory
- MemLiner: A remote-memory-friend runtime system
- Mako: A low-pause, high-throughput evacuating collector for memory-disaggregated datacenters
- Dorylus: Affordable and scalable GNN training with scale-out CPU servers and thousands of serverless threads
- JPortal: Java Virtual Machine support for Intel PT
- Chianina: An evolving graph system for fully context- and flow-sensitive static analysis
- Semeru: A memory-disaggregated managed runtime
- Reducto: an on-camera frame filtering system for
efficient video analytics
- Grapple: A graph system for finite state property checking for very large systems
- RStream: A
single-machine, disk-based graph mining system
- Graspan: A
disk-based parallel graph system for dynamic transitive closure
computation
- Facade: A compiler and
runtime system for efficient Big Data systems
- Statistical
debugging based on abstraction refinement
- PerfBlower:
A performance problem amplification framework and an instrumentation
specification language
- GraphQ:
Analytical query processing based on abstraction refinement over very
large graphs
- Resurrector: A new object lifetime profiler
- Demand-driven
context-sensitive alias analysis for Java
- LeakChaser: A Java memory leak detector
- AJANA: A framework for source-code-level
analysis of AspectJ software
Why I Believe
I am a Protestant Christian. I regularly attend and help with several campus fellowships at both UCLA and UC Irvine.
- How can one be a CS professor and a Christian? (collected by George Varghese)
Contact Information
harryxu at cs dot ucla dot edu
http://www.cs.ucla.edu/~harryxu
Office:
Engineering VI, Room 496A
Computer Science Department
BOX 951596, UCLA
Los Angeles, CA 90095-1596
Work: 310-794-7145
Last updated: June 8, 2018
Maintained by harryxu at cs
dot cs dot ucla dot edu (replace dot with .)