Biography
I'm a Senior Computer Science student @ShanghaiTech. I'm
honored to be under polished by three system related Professors Foo, Shu and Toast no matter for my research career, my course journey
and high performance cluster competitions.
case class DRAM(tag: String, taste: String, percentage: Int)
val OnCrash = List(
DRAM("HPC Training", "Nice", 40),
DRAM("NVRAM Hooligon", "Damm Hard" 41),
DRAM("Compiler TA", "Thankless", 18),
DRAM("Course Work", "Indifferent", 1))
val CoffeeTime =
for (OnCrash <- Machine if Machine.percentage.delta + Machine.mt19937ar.rand >= 1
&& strcmp(Machine.taste,"So So") >0 && Machine.percentage--) //lazy eval
yield Machine.tag
CoffeeTime.foreach(println)
Skills
Education
BEing in Computer Science and Engineering
Shanghai, China
September 2018 - July 2022
Ph.D. Student in Computer Science and Engineering
Santa Cruz, California
October 2022 - July 2028
Industrial Experiences
Serving for HFT team by HPC automated DevOps stack
Shanghai, China
June 2020 - July 2020
- High Frequency Trade Order Book simulation applying Linear.Regression Method.
- Applied salt and jinja to automate scheduling of jobs and assigning affinity of cpu cores in Linux DevOps.
- Applied gobidng of gobpf to try IOVisor stuff.
Leader
September 2019 - November 2021
Established a team of 16 undergraduate students for HPC related research. Navigated the first
research funding from Inspur and Nvidia. Continually preparing and aiming for the top-tier student
supercomputing competitions, including ASC,ISC
and SC.
- Optimized a parallel quantum computer simulater QuEST by reducing the cache-miss rate and introducing AVX2/AVX512 vectorization for CPU part and adopting NCCL comunication infro for GPU part.
- Compiled and distributed climate simulator CESM on Azure with automated scheduling.
- Optimized a Profiler of MPI alltoallv called collective_profiler by a Nvidia Architect Geoffroy Vallee.
- Added a parallel I/O Middleware for WRF which increase around 30% of the performance.
- Applied card binder on 8 card A100 with NVLink to quantum espresso and HPL/HPCG, Standard_ND96asr_v4 which top the board.
Research Interests
- Computer System (Arch)
- Blockchain
- Reinforcement Learning
- High Performance Computing
- GPGPU
- Symbolic Execution
- Programming Language
Research Experiences
- Researching Adversarial Sample Detection for Deep Neural Network using foolbox and IBM-ART to improve accuracy of object recognition to prevent impostors from hacking into systems.
- Researching MOVE language in Libra currency source code to protect against Arithmetic Overflow, Timestamp Dependency.
- Understanding the real access mechanism of Memory Mode Optane Memory and XPBuffer by reverse-engineering methods.
- Investigating Order-Dependent Flaky Tests.
Publications
Attack as Defense: Characterizing Adversarial Examples using Robustness
Highlighted Projects
-
The value of VChain comes from the application of driver-assisted decision making
and the
innovation of the business model, the former relying on blockchain security and
computation
without latency, and the latter as a result of data aggregation and movement.
- In one scope, multiple car sensors and their data transactions are trusted by the blockchain created by CPChain.
- Artificial intelligence decision making, based on IoT big data. Heterogeneous data interoperability, resulting in big data aggregation.
- Interaction and application of data assets, data assets are leveraging the verifiable characteristics of the main chain to benefit emerging business models such as insurance valuation.
