Kai Londenberg
Senior Machine Learning Engineer at Bioptimus
Burgdorf, Lower Saxony, Germany
Summary
🤩
Rockstar🎓
Top SchoolKai Londenberg is a Senior Machine Learning Engineer with over two decades of professional experience and a 12-year focus on AI, currently accelerating multimodal foundation models for medical and biotech applications at Bioptimus. He has a strong systems and performance background from Meta and significant open-source impact—optimizing PyTorch’s Inductor Cutlass backend for GEMM and bmm operations, multithreaded precompilation and improved autotuning. Comfortable across Python, CUDA, C/C++ and HPC stacks, he combines deep learning, Bayesian statistics and GPU engineering to push models from research into production. Previously he led data science teams at Volkswagen and Searchmetrics, and earlier built startups and tooling, reflecting a blend of product-minded engineering and research rigor. Based in Lower Saxony, Germany, he’s a lifelong learner who intentionally bridges low-level performance work with high-level generative AI systems for real-world impact.
12 years of coding experience
23 years of employment as a software developer
Various, Various at CourseraUdacityMathematics / Computing Science, Mathematics / Computing Science at Universität Hannover
German, English
Stackoverflow
131reputation
9kreached
4answers
1question
Github Skills (21)
auto-tuning10
pytorch10
back-end-development10
python10
gemfire10
gpu10
performance-optimization10
cuda10
c-language9
tensor9
cprogramming-language9
deep-learning8
triton8
machine-learning8
debug-symbols6
Programming languages (6)
Github contributions (5)
pytorch/pytorch
Apr 2017 - Mar 2025
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Role in this project:
Back-end Developer & Performance Engineer
Back-end Developer & Performance EngineerContributions:236 reviews, 148 PRs, 2211 pushes in 8 years
Contributions summary:Kai primarily focused on optimizing the Inductor Cutlass backend within the PyTorch project, specifically targeting GEMM (General Matrix Multiply) operations. Their contributions included upgrading and adapting the backend to newer Cutlass versions, implementing and refining GEMM shape padding, and adding support for bmm (batch matrix multiply) operations. Further, they incorporated performance enhancements with multithreaded precompilation and improvements to autotuning, which included graceful handling of slow or problematic kernels and improved logging of autotuning results.
pythongpu-accelerationdeep-learninggpunumpy
kadeng/pytorch
Feb 2023 - Mar 2025
Tensors and Dynamic neural networks in Python with strong GPU acceleration
Contributions:38 pushes, 10 branches in 2 years 1 month
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.