Lars Kuehmichel
Doctoral Researcher at Technische Universität Dortmund
Dortmund, North Rhine-Westphalia, Germany
Summary
👤
Senior🎓
Top SchoolLars Kuehmichel is a computational statistics PhD researcher at TU Dortmund with eight years of experience building deep learning and Bayesian solutions, particularly in generative modeling, domain generalization, and high-performance computing. He combines a strong physics background (BSc/MSc Heidelberg) with hands-on ML engineering, contributing to major open-source projects like Keras and the FrEIA invertible networks framework where he implemented spline-based modules and numerical ops. His work spans both research—evidenced by OpenReview publications—and production-ready tooling, improving core math operations and backend robustness. Known for bridging rigorous Bayesian thinking with practical neural architecture design, he also brings experience from a visiting research stint at RPI, reflecting an international collaborative profile.
8 years of coding experience
Master of Science - MS, Physics, Master of Science - MS, Physics at Ruprecht-Karls-Universität HeidelbergBachelor of Science - BS, Physics, Bachelor of Science - BS, Physics at Heidelberg University
German, English, French
Github Skills (11)
web-framework10
pytorch10
machine-learning10
tensor10
deep-learning10
tensorflow10
python10
numpy9
neural-network9
jax8
data-science7
Programming languages (7)
Github contributions (5)
vislearn/FrEIA
Sep 2022 - Jan 2023
Framework for Easily Invertible Architectures
Role in this project:
ML Engineer
ML EngineerContributions:8 reviews, 16 commits, 5 PRs in 4 months
Contributions summary:Lars contributed to the `FrEIA` framework by implementing and modifying modules related to invertible neural networks. Their primary focus was on developing and refining spline-based modules, including the addition of linear and rational-quadratic splines and the reparameterization of `BinnedSpline`. They also worked on improvements to the `ActNorm` module and included distributions outlines, indicating a focus on the mathematical and architectural aspects of the framework.
architecturesopenflowframeworkdata-mesharchitecture
keras-team/keras
Feb 2024 - Feb 2025
Deep Learning for humans
Role in this project:ML Engineer
ML EngineerContributions:3 reviews, 5 PRs, 28 comments in 11 months
Contributions summary:Lars contributed to the Keras library by implementing and modifying core mathematical operations. They improved `keras.ops.isclose`, adding support for tolerances and backend implementations. They also added `keras.ops.dtype` and `keras.ops.searchsorted`, expanding the available operations within the library. Furthermore, they fixed an unbound local error in the Torch backend related to training logs.
pythondata-sciencedeep-learningneural-networksmachine-learning
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.