Bjarke Roune
Principal Compiler Engineer at NVIDIA
Mountain View, California, United States
Summary
👤
Senior🎓
Top SchoolBjarke Roune is a Principal Compiler Engineer based in Mountain View with 13 years of experience specializing in compilers, hardware-software co-design, ML infrastructure and performance optimization. He has driven compiler and ML library development for Google's TPUv3, led a software-only power management change that unlocked 23% more hardware flops, and contributed to production ML compilers at Amazon (SageMaker Neo) and NVIDIA. His work spans research-grade foundations—PhD-level algebraic computation and postdoctoral research—to practical backend engineering, including improving HLO text formatting and profiling in the openxla XLA project to aid debugging and profiling. Known for translating deep theory into high-impact engineering, he blends academic rigor with a track record of shipping compiler features that materially improve hardware utilization.
13 years of coding experience
10 years of employment as a software developer
Doctor of Philosophy (PhD) Computer Science, Doctor of Philosophy (PhD) Computer Science at Aarhus UniversityGymnasium (High School), Gymnasium (High School) at Haderslev Katedralskole
English, Danish
Stackoverflow
3,667reputation
180kreached
19answers
15questions
Github Skills (12)
profiling10
cprogramming-language9
c-language9
debug8
debugging8
nested6
unit-testing6
assembly6
compiler-errors6
visual-studio6
compiler-construction6
testing6
Programming languages (4)
Github contributions (5)
openxla/xla
Jan 2017 - Jun 2019
A machine learning compiler for GPUs, CPUs, and ML accelerators
Role in this project:
Back-end Developer
Back-end DeveloperContributions:33 commits in 2 years 5 months
Contributions summary:Bjarke's contributions focused on improving the HLO text format printing functionality. Their work involved modifications to the code related to HLO instruction formatting, including enhancements to display information such as dilation and index values. These changes likely aimed to enhance debugging and readability of the HLO representation. They also added sum-across-opcodes report for HLO profiling.
compilercommunity-drivenmachine-learningmodular
broune/mathic
Jul 2012 - Aug 2013
C++ library of symbolic algebra data structures for use in Groebner basis computation.
Contributions:49 commits in 1 year 1 month
mathematicsbasissymbolic-algebracascomputation
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.