James Hegeman
Software Engineer, AI Infra at Facebook
Sunnyvale, California, United States
Summary
🤩
Rockstar🎓
Top SchoolJames Hegeman is a software engineer specializing in ML/AI infrastructure with eight years of hands-on experience building low-level software for specialized hardware and compiler backends. Currently a Research Scientist / Software Engineer at Facebook AI Infra, he focuses on NN compilation, backend JIT work, and software/hardware co-design for inference on custom silicon. He contributed to the PyTorch Glow compiler—adding instruction support, test automation, and bug fixes—demonstrating a strong blend of compiler engineering and rigorous validation. Prior roles include ML compiler work at Waymo and extensive academic research and cluster administration during his graduate studies, reflecting both production and research chops. Trained in mathematics and computer science, he brings a methodical, analytically driven approach to performance-critical ML systems and a knack for improving test infrastructure that often goes unnoticed.
8 years of coding experience
6 years of employment as a software developer
Master of Arts (MA), Mathematics, Master of Arts (MA), Mathematics at University of Wisconsin-MadisonBachelor of Science (BS), Mathematics, Bachelor of Science (BS), Mathematics at California Institute of TechnologyMaster of Computer Science, PhD Candidate (ABD), Computer Science, Master of Computer Science, PhD Candidate (ABD), Computer Science at University of Iowa
Github Skills (7)
c-language10
cprogramming-language10
jit10
unit-test10
compile10
llvm10
machine-learning8
Programming languages (2)
Github contributions (4)
pytorch/glow
Jan 2018 - Nov 2018
Compiler for Neural Network hardware accelerators
Role in this project:
Back-end Developer & Test Automation Engineer
Back-end Developer & Test Automation EngineerContributions:79 commits, 9 PRs, 3 pushes in 10 months
Contributions summary:James primarily contributed to the back-end functionality of the Glow compiler for neural network hardware accelerators. They implemented support for the Tanh, Sigmoid, ElementMin, ElementSelect, and other instructions within the JIT (Just-In-Time) compilation backend. Furthermore, the user added test functions and automated unit tests for correctness, ensuring the functionality of the implemented instructions, including the development of a test generation tool. They also fixed bugs related to element-wise operations and provided improvements for the test infrastructure.
hardware-acceleratorscompilerneural-networkacceleratorshardware
hegemanjwh2/glow
May 2018 - May 2018
Contributions:3 pushes, 1 branch in 4 days
hardware-acceleratorscompilerneural-networkacceleratorshardware
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.