Johannes Schönberger
Research Scientist at Meta
Zurich, Zurich, Switzerland
Summary
🤩
RockstarJohannes Schönberger is a research scientist and experienced engineer based in Zurich with 14 years of software development and research experience across computer vision, numerical optimization, and scientific computing. He contributes deeply to foundational open-source projects—Ceres Solver, NumPy, scikit-learn, COLMAP and vcpkg—often focusing on mathematically sensitive backend work like automatic differentiation, robust estimators (RANSAC), and quaternion-to-matrix conversions. Johannes is comfortable across the stack, from low-level C++ build and compiler fixes to Python API design and testing, and he has a track record of improving numerical robustness, reducing memory footprint, and resolving subtle race conditions. His contributions to high-profile projects indicate a blend of research rigor and production engineering, making complex algorithms reliable and maintainable for broad use. Colleagues value his attention to edge cases and cross-platform compatibility—skills that repeatedly surface in fixes for compilers, geospatial buffering, and multi-view reconstruction pipelines.
14 years of coding experience
Github Skills (52)
unit-testing10
c-language10
lib10
package-management10
python10
image-processing10
signal-processing10
scikit10
data-manipulation10
machine-learning10
code-library10
cmake10
numpy10
automatic-differentiation10
geometric-algorithms10
Programming languages (15)
Github contributions (5)
microsoft/HoloLensForCV
Jan 2018 - Aug 2020
Sample code and documentation for using the Microsoft HoloLens for Computer Vision research
Role in this project:
Full-stack Developer
Full-stack DeveloperContributions:40 commits, 11 PRs, 15 pushes in 2 years 7 months
Contributions summary:Johannes primarily focused on improving the "Recorder" tool within the HoloLensForCV project. Their contributions involved fixing race conditions, addressing bugs, and adding enhancements to the recorder application's functionality. They also worked on integrating camera view and projection transforms, indicating involvement in the core data capture and processing pipeline. Furthermore, the user made changes to the build system, updating NuGet packages and SDK versions.
visiondeep-learningcomputer-visionmicrosofthololens
colmap/colmap
Apr 2016 - Jan 2023
COLMAP - Structure-from-Motion and Multi-View Stereo
Role in this project:Back-end Developer
Back-end DeveloperContributions:25 releases, 711 reviews, 1375 commits in 6 years 10 months
Contributions summary:Johannes contributed to the COLMAP project by implementing a method to convert quaternions to a rotation matrix within the read_model.py script. Additionally, the user fixed a bug by adding missing SQL statement finalization in src/base/database.cc. Further contributions included the addition of the automatic reconstruction pipeline, and several improvements to the function of point cloud visualizer.
structure-from-motionreconstructionhomographymulti-viewstereo
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.