Christian Puhrsch
Software Engineer at Meta
Seattle, Washington, United States
Summary
🤩
Rockstar🎓
Top SchoolChristian Puhrsch is a software engineer based in Seattle with 11 years of experience building high-performance back-end systems, currently at Meta. He combines a strong academic foundation in computer science and statistics (University of Toronto) and data science (NYU) with practical expertise in ML infrastructure, contributing to major open-source projects like fastText and PyTorch. His work focuses on performance-sensitive components—sparse tensor layouts, optimized conversion routines, and efficient data loaders—that directly improve training and inference throughput. As an adjunct instructor in deep learning, he also translates research ideas into usable code and educational material. Colleagues value his attention to low-level detail and his knack for making complex numerical code both faster and more maintainable.
11 years of coding experience
Master in Science, Data Science, Master in Science, Data Science at New York UniversityHonours Bachelor of Science with Distinction, Computer Science and Statistics, Honours Bachelor of Science with Distinction, Computer Science and Statistics at University of Toronto
English, German
Github Skills (27)
pytorch10
c-language10
lib10
operation10
python10
tensorrt10
matrix-multiplication10
sparse-matrix10
code-library10
datasets10
data-loading10
tensorflow10
gpu10
performance-optimization10
nlp10
Programming languages (8)
Github contributions (5)
facebookresearch/fastText
May 2017 - Jan 2018
Library for fast text representation and classification.
Role in this project:
Back-end Developer
Back-end DeveloperContributions:1 release, 79 commits, 44 PRs in 8 months
Contributions summary:Christian primarily contributed to the core functionality of the `fasttext` library. Their work involved implementing performance optimizations by avoiding the use of `std::thread` for single-threaded operations, and adding file format versioning and checking. They also added sentence vector calculations and print functions. This suggests a focus on improving the efficiency, reliability, and usability of the library.
classifierbertword-embeddingsword2vectopic-modeling
pytorch/pytorch
Feb 2018 - Dec 2022
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:Back-end Developer
Back-end DeveloperContributions:1653 reviews, 227 commits, 416 PRs in 4 years 11 months
Contributions summary:Christian's contributions focused on enhancing the PyTorch framework, specifically concerning sparse tensors. The user implemented and optimized conversion routines between sparse and strided layouts, including the addition of a new SparseBsr layout. The user also worked on a variety of operator implementations and enhancements, particularly for bmm with NestedTensors, and supporting bfloat16 for it. Furthermore, they introduced a primitive for int8@int8 -> int32 matrix multiplication.
pythongpu-accelerationdeep-learninggpunumpy
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.