Thorsten Kurth
Senior Software Engineer at NVIDIA
Zurich, Zurich, Switzerland
Summary
🤩
Rockstar🎓
Top SchoolThorsten Kurth is a Senior Software Engineer and Gordon Bell Prize awardee with 11+ years driving performance-critical scientific and AI software, currently working at NVIDIA from Zurich. He specializes in performance optimization, parallel and distributed deep learning at extreme scale—co-leading the first deep learning application to exceed one ExaOp/s on Summit—and has led DOE NESAP efforts to accelerate AI codes on supercomputers. His background spans algorithm development, accelerator programming models, CUDA optimizations, I/O tuning and statistical analysis of multi-terabyte datasets, and he contributed high-impact backend work to NVIDIA/DALI (numpy reader, parallel GPU chunking and math kernels). Combining a PhD in theoretical physics with hands-on HPC and production GPU engineering, he thrives at the intersection of research and deployable systems and is seeking project-lead or expert roles in performance-critical AI, especially in technology or automotive domains.
11 years of coding experience
7 years of employment as a software developer
Diplom, Physik, Sehr gut, Diplom, Physik, Sehr gut at Bergische Universität Wuppertal
English, German, French
Stackoverflow
13reputation
399reached
0answers
2questions
Github Skills (17)
python10
numpy10
gpu10
data-processing10
c-language9
machine-learning9
cprogramming-language9
deep-learning8
image-processing8
data-augmentation7
nlp6
c996
nltk6
gensim6
gcc6
Programming languages (8)
Github contributions (5)
NVIDIA/DALI
Apr 2020 - Jul 2021
A GPU-accelerated library containing highly optimized building blocks and an execution engine for data processing to accelerate deep learning training and inference applications.
Role in this project:
Back-end Developer
Back-end DeveloperContributions:15 reviews, 7 commits, 9 PRs in 1 year 3 months
Contributions summary:Thorsten contributed significantly to the NVIDIA DALI (Data Loading Library) project by implementing features and improving the existing functionality of the numpy reader operator. Their work involved adding the numpy reader, supporting file list arguments, and enabling parallel chunking for GPU operations. Furthermore, they added exp and log math functions to the project. The user also refactored the code by renaming file loaders and disabling buffer registration for the GPU variant.
pythontrainingpaddlegpu-accelerationbuilding-blocks
azrael417/latdata_analyze
Jan 2015 - May 2015
Contributions:479 pushes in 4 months
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.