Chengfan Jia
Jiading District, Shanghai, China
Summary
🤩
Rockstar🎓
Top SchoolChengfan Jia is an engineer with 11 years' experience specializing in AI compilers and performance optimization, currently at Tencent after leading device-focused algorithm research at NIO. He is an active Apache TVM committer and contributor to the high-profile TVM project, with hands-on work optimizing CUDA device APIs, GPU verification passes, and runtime performance. His background spans deploying optimized models across ARM, NVIDIA DRIVE AGX and custom AI ASICs, with deep expertise in graph optimization, quantization and op fusion. Trained in computer architecture (USTC) and with industry experience at Alibaba and Apple, he blends research-grade algorithmic rigour with production delivery for edge and cloud ML workloads. Notably, he has helped put autonomous-driving models onto vehicles by integrating end-to-end compiling and deployment pipelines.
11 years of coding experience
2 years of employment as a software developer
Master of Science (M.S.) Computer Architecture, Master of Science (M.S.) Computer Architecture at University of Science and Technology of ChinaHigh School, High School at Zhejiang Wenling High SchoolBachelor of Engineering (B.Eng.) Communication Engineering, Bachelor of Engineering (B.Eng.) Communication Engineering at Chang'an University
Chinese, English
Github Skills (10)
cuda10
compiler10
performance-monitor10
compiler-compiler10
performance-analysis10
python9
cprogramming-language9
c-language9
deep-learning8
machine-learning7
Programming languages (7)
Github contributions (5)
apache/tvm
Jun 2020 - Dec 2021
Open Machine Learning Compiler Framework
Role in this project:
Backend Developer & Performance Engineer
Backend Developer & Performance EngineerContributions:347 reviews, 31 commits, 100 PRs in 1 year 5 months
Contributions summary:Chengfan contributed to the CUDA device API and VerifyGPUCode pass, including adding device APIs for CUDA and vectorization checks. The changes primarily focused on the src/tir/analysis/verify_gpu_code.cc file, suggesting involvement in GPU code verification and optimization. The user's work also extends to the runtime/cuda and tests/python/unittest directories, reflecting efforts to ensure code correctness and improve performance on CUDA-enabled devices within the TVM compiler stack.
compilermachine-learningtensordeep-learninggpu
jcf94/tvm
Oct 2019 - Aug 2021
Open deep learning compiler stack for cpu, gpu and specialized accelerators
Contributions:130 commits, 1 PR, 299 pushes in 1 year 10 months
compilerdeep-learninggpu
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.