Dehao Chen - Team Lead Manager at Waymo

Dehao Chen

Team Lead Manager at Waymo

Mountain View, California, United States

Join Prog.AI to see contacts

Summary

🤩

Rockstar

🎓

Top School

Dehao Chen is a senior compiler and ML performance leader with 14 years of experience optimizing compilers and hardware-software co-design for large-scale ML workloads. He has deep expertise across XLA, LLVM, GCC and AutoFDO, driving sample-based PGO, debug-info improvements, and codegen optimizations that materially improve runtime and profiling fidelity. As Waymo's Team Lead Manager he now directs compute software efforts—timing, system tools and TPU-focused ML performance—translating research techniques into production training and inference wins. His decade at Google included spearheading MLPerf submissions and scaling TPU workloads, and his open-source contributions to LLVM/Clang and AutoFDO reflect a practical commitment to tooling that benefits broad compiler ecosystems. Notably, his work frequently targets the subtle intersection of debug information and profiling accuracy, a less-visible lever that yields significant performance and observability gains.

14 years of coding experience

10 years of employment as a software developer

Bachelor, Computer Science, Bachelor, Computer Science at Huazhong University of Science and Technology

Ph.D, Computer Science - Compiler technology, Ph.D, Computer Science - Compiler technology at Tsinghua University

English

Github Skills (28)

debugging10

debug10

c-language10

dwarf10

llvm10

testing10

compiler-development10

gcov10

c1110

c1710

compiler-compiler10

performance-optimization10

compiler10

cprogramming-language10

optimization10

Programming languages (6)

C++CLLVMMakefileAssemblyPython

Github contributions (5)

google/autofdo

Apr 2014 - Apr 2018

AutoFDO

Role in this project:

Back-end Developer & Performance Engineer

Contributions:7 releases, 52 commits, 31 PRs in 4 years 1 month

Contributions summary:Dehao primarily contributed to the performance and functionality of the `autofdo` project, which involves automatic feedback-directed optimization. Their work included implementing new profile tools like `dump_gcov`, `profile_merger`, and `profile_update`, enhancing the project's capabilities. They also focused on optimizing code, as seen in the performance and memory optimization commit, improving the efficiency of the profiling process. Additionally, the user fixed bugs and integrated updates to the project's codebase, further contributing to its stability and performance.

react

apple/swift-llvm

Nov 2015 - Aug 2017

Role in this project:

Back-end Developer

Contributions:10 commits in 1 year 9 months

Contributions summary:Dehao primarily contributed to improving the debug information emitted for profile-guided optimization (PGO) in the Swift-LLVM project. Their work focused on adding extra debug information to the compiler's output, specifically including the start line, linkage name, and status of subprograms, to enhance the accuracy of sample PGO profile collection. These changes involved modifications to the target options, the DwarfUnit and DwarfDebug classes, and integration into the test suite. The overall impact of the changes was a size increase in the compiled binaries while providing more detailed profiling data.

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.

Request Free Trial