Hao Zhang
System Architect Scientist at 4Paradigm 第四范式
Singapore
Summary
🤩
Rockstar🎓
Top SchoolHao Zhang is a system architect scientist based in Singapore with eight years of experience building high-performance distributed systems and ML infrastructure. He holds a PhD from NUS where he researched in-memory and distributed database architectures and has published in VLDB and ICDE. At 4Paradigm he architects production ML platforms and previously improved deployment and build systems for OpenMLDB, cutting Docker image sizes and streamlining zetasql integration. His low-level contributions to the widely used TVM compiler—especially VTA runtime, device annotations, and quantization support—showcase expertise at the intersection of compilers, hardware accelerators, and ML. Comfortable across research and production, he combines deep systems thinking with practical DevOps and backend engineering to optimize performance on specialized hardware.
8 years of coding experience
4 years of employment as a software developer
Bachelor’s Degree Computer Science, Bachelor’s Degree Computer Science at Harbin Institute of TechnologyDoctor of Philosophy (Ph.D.) Computer Science, Doctor of Philosophy (Ph.D.) Computer Science at National University of Singapore
English, Chinese
Github Skills (20)
performance-monitor10
c-language10
docker10
machine-learning10
bash10
dockers10
compiler-compiler10
performance-analysis10
deep-learning10
gpu10
compiler10
cprogramming-language10
cicd9
python9
build-automation9
Programming languages (5)
Github contributions (5)
4paradigm/OpenMLDB
Nov 2021 - Jan 2023
OpenMLDB is an open-source machine learning database that provides a feature platform computing consistent features for training and inference.
Role in this project:
Back-end & DevOps Engineer
Back-end & DevOps EngineerContributions:1 release, 498 reviews, 58 commits in 1 year 2 months
Contributions summary:Hao primarily focused on improving the project's build and deployment process. They reduced the size of the demo Docker image by optimizing dependencies and removing unnecessary components. The user also enhanced the project by adding the `OPTIONS` parameter to the `DEPLOY` statement and incorporating a pre-built zetasql library. Furthermore, they made changes to the build scripts and updated the configuration for the deployment of the system.
machine-learning-databasetrainingfeature-storein-memory-databasefeature-extraction
apache/tvm
Apr 2020 - Apr 2021
Open deep learning compiler stack for cpu, gpu and specialized accelerators
Role in this project:Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:4 reviews, 6 commits, 14 PRs in 11 months
Contributions summary:Hao primarily contributed to the TVM compiler stack, focusing on low-level runtime and compilation aspects, including VTA (Versatile Tensor Accelerator) support. Their work involved modifying runtime components for memory management, adding OpenCL file type support for linting, and implementing device annotation changes related to Relay. Additionally, the user integrated quantization support for ALU-only operations and added device annotation support in graphpack. These contributions improve the efficiency and functionality of the compiler for deep learning tasks, especially for specialized hardware like the VTA.
metalvulkancompilertensoropencl
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.