Xiaoyang Wu

Senior GPU Compiler Engineer - ISS Developer @ Qualcomm

About

I received my UCSD MS degree in Computer Science in 2023. And started working in Qualcomm as GPU Compiler Engineer after graduation.In the previous years I joined Tencent(one of the biggest companies in China) and Tiktok as SDE intern. I worked mainly on the development of applications and toolset related to distributed data processing and machine learning algorithm(e.g. applying Diffusion Model to video-generation apps). During my work experience in Qualcomm, I independently participated in the development of high-level GPU emulator at both LLVM IR level and Machine IR level, helping in catching correctness issues and performance tuning.I can utilize my skills and experience in programming languages such as Java, Python, C/C++, and is familiar with machine learning tools and techniques, and have solid foundation on data structures, algorithms, system design, backend development and machine learning engineering.I am open to opportunities in a variety of fields and if you or your company are looking for a talented Software Engineer, please feel free to reach out to me..

Country

United States

City

San Diego

Industry

Computer Hardware

Skill

LLVM, GPU, Compilers, Docker, Compilation, Cloud Applications, Machine Learning, Cython, Distributed Algorithms, Python (Programming Language), Data Structures, JavaScript, Linux, SQL, Git, English, Engineering, Problem Solving, Java, Python

Experience

Senior GPU Compiler Engineer - ISS Developer

Qualcomm

2023-6 - Present · 3 yrs 4 mos

San Diego, CA

• GPU Simulator: Participated in developing GPU simulator at LLVM IR level and Machine IR level on OpenCL/Vulkan/DX11/DX12 shaders. Providing stats and analysis for compiler verification, performance tuning, new-feature development.

Software Engineer Intern

TikTok

2022-6 - 2022-9 · 4 mos

Mountain View, California, United States

• Developed distributed data processing infrastructure in Python&Cython. It compiles serialized for-loops and execute them in distributed environment while being able to resume from break points when program restarts. Helped in improving the efficiency of python-based data processing applications at least 4x faster, more scalable and more reliable in distributed environment. • Lead a team of 5 to develop an application which produces MV given music&lyrics as inputs. Utilized CLIP-guided Diffusion model to implement the architecture and Real-ESRGAN to apply super-resolution effects onto frames. Utililzed above-mentioned distributed toolset to scale the application onto 32 nodes, enabling it to process 200 million musics within 2 weeks.

Software Engineer Intern

Tencent

2021-5 - 2021-9 · 5 mos

Shanghai, China

• Developed a cover-image generator which takes videos as input and generates the most suitable cover-image for the video. Utilized PaddleOCR to detect texts within frames, trained and optimized models using GoogleNet as the model framework to predict the scores of each frame with respect to its various qualities. • Deployed the above-mentioned application onto Tencent Could Service and designed RESTful API skeleton for user access. • Helped in dealing with model compatibility issues. Designed/modified compilers to convert between various DL model description formats(e.g. caffe-to-onnx) and designed/modified dockerfile to create compatible environments.