Jiahao Luo
Software Engineer at Huawei
Pudong, Shanghai, China
Summary
🤩
Rockstar🎓
Top SchoolJiahao Luo is a software engineer with eight years of experience blending strong statistical foundations (B.S. in Statistics, M.A. from Columbia) with systems and ML engineering across industry leaders like ByteDance and Huawei. He specializes in low-level performance work—CUDA/Metal backends, GPU kernel optimization, and in-place tensor ops—demonstrated by notable contributions to high-profile open-source projects such as ggml, llama.cpp, and torch-mlir. Comfortable across C/C++, Python, Go, and kernel/OS spaces, he bridges ML research and production by lowering PyTorch ops to MLIR and improving inference primitives like RoPE, Alibi, GELU, softmax and RMSNorm. Colleagues value his analytical rigor and persistence, rooted in math contests and rigorous academics, which he pairs with practical teamwork from diverse project and extracurricular experience.
7 years of coding experience
1 year of employment as a software developer
Master of Arts - MA, Statistics, 4.00/4.00, Master of Arts - MA, Statistics, 4.00/4.00 at Columbia University in the City of New YorkBachelor's degree, Statistics, 3.84/4.00, Bachelor's degree, Statistics, 3.84/4.00 at Zhejiang University
Github Skills (28)
pytorch10
c-language10
operation10
tensorrt10
gpu-programming10
machine-learning10
large-language-models10
mlr10
llm10
metal10
compiler-compiler10
gm10
tensorflow10
performance-optimization10
cuda10
Programming languages (6)
Github contributions (5)
ggml-org/ggml
May 2023 - May 2024
Tensor library for machine learning
Role in this project:
ML Engineer
ML EngineerContributions:1 review, 16 PRs, 8 comments in 1 year
Contributions summary:Jiahao made significant contributions to the ggml library, which is designed for machine learning applications. Their work involved implementing and expanding the API for inplace operations, allowing for in-place modifications of tensors, thus optimizing memory usage. They also added support for ChatGLM-style RoPE (Rotary Position Embedding) within the CUDA implementation, integrating techniques used in large language models. Furthermore, the user added new CUDA kernels, including a GELU (Gaussian Error Linear Unit) activation function and support for an Alibi (Attention with Linear Biases) operator.
deep-learningllmmachine-learningtensor-libraryneural-network
llvm/torch-mlir
Dec 2022 - Jan 2023
The Torch-MLIR project aims to provide first class support from the PyTorch ecosystem to the MLIR ecosystem.
Role in this project:ML Engineer
ML EngineerContributions:38 reviews, 8 commits, 20 PRs in 1 month
Contributions summary:Jiahao contributed significantly to the Torch-MLIR project, focusing on extending the support for PyTorch operations within the MLIR ecosystem. Their work included adding and decomposing `aten.std.correction` and enhancing e2e testing for logical and bitwise operations. They also contributed to the MHLO conversion process, specifically handling `aten.copy`, `aten.rsqrt`, and `aten.sigmoid` operations, demonstrating a focus on lowering PyTorch operations to the MHLO dialect.
pytorchmlirtorchcompilerecosystem
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.