Johan Von Forstner
AI R&D Architect at PARADOX CAT GmbH
Munich, Bavaria, Germany
Summary
🤩
Rockstar🎓
Top SchoolJohan Von Forstner is an AI R&D Architect with 13 years of experience who leads and mentors teams building in-cabin perception and human–machine interaction solutions for automotive customers such as BMW. Holding a PhD in Space Physics, he brings deep expertise in complex sensor data processing—from spaceborne instruments to 60 GHz radar and UWB cabin sensors—and a proven track record of turning research into optimized embedded deployments. He is fluent in Python and modern ML stacks (PyTorch, TensorFlow, JAX) and bridges the full ML engineering workflow from sensor choice and calibration to deployment and CUDA acceleration. An active open-source contributor, he has improved foundational libraries like matplotlib and fastai and ensured robustness in SunPy through testing and QA work. Based in Munich, he pairs academic rigor (25 peer-reviewed papers) with pragmatic product focus, and maintains an EV mapping side project that reflects his interest in real-world mobility applications.
13 years of coding experience
8 years of employment as a software developer
Dr. rer. nat. (PhD) Physics, Dr. rer. nat. (PhD) Physics at Kiel University
German, English, French, Norwegian
Stackoverflow
147reputation
5kreached
4answers
3questions
Github Skills (34)
data-visualizations10
pytorch10
pytest10
python10
matplotlib10
testing10
machine-learning10
data-visualisation10
plot10
astrophysics10
java10
android-library10
javas10
solar10
androidx10
Programming languages (17)
Github contributions (5)
opacapp/multiline-collapsingtoolbar
Dec 2015 - Jun 2019
A modified CollapsingToolbarLayout that can deal with multiline titles
Role in this project:
Mobile Developer (Android)
Mobile Developer (Android)Contributions:57 commits, 2 PRs, 39 pushes in 3 years 7 months
Contributions summary:Johan primarily contributed to the Android-specific collapsing toolbar library, `multiline-collapsingtoolbar`. They removed unnecessary imports and documented changes from the original Support Library code, indicating a focus on code cleanliness and maintainability. Subsequent commits involved updating the library to support changes from the Android Support Library, including fixes for compatibility and functionality, demonstrating a deep understanding of the library's internal workings and Android development principles. The user also added a demo app and made various improvements and bug fixes to the library.
titlebarcollapsingtoolbarlayoutdealandroidmultiline
fastai/fastai
Aug 2021 - May 2022
The fastai deep learning library
Role in this project:ML Engineer
ML EngineerContributions:5 commits, 5 PRs, 6 comments in 9 months
Contributions summary:Johan primarily focused on improving the `fastai` deep learning library. Their contributions included fixing a bug related to the `crop_pad` image augmentation functionality, ensuring correct behavior in cropping and padding. They also addressed issues with the `plot_top_losses` function, particularly for models with multiple outputs, and enhanced the AzureML callback by adding logging improvements, including logging to the parent run in pipelines. Additionally, the user corrected an issue related to `pin_memory=True` in DataLoader and fixed the `torch.jit.script` compatibility of the TimmBody.
pytorchpythondeep-learninggpumachine-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.