Marek Kowalski
Principal Scientist at Microsoft
Warsaw, Masovian Voivodeship, Poland
Summary
👤
Senior🎓
Top SchoolMarek Kowalski is a Principal Scientist based in Warsaw with over a decade of expertise in computer vision and machine learning, specializing in facial and body understanding and both traditional and ML-driven 3D reconstruction. At Microsoft he has progressed from research intern to principal scientist, turning research into product-ready demos and shipping systems that stream real-time point clouds to devices like HoloLens. He is a published and patented researcher (ECCV/ICCV/CVPR and eight patent applications) who also codes end-to-end—authoring Unity-based HoloLens apps and Python tools for 3D face swapping and multi-sensor LiveScan3D. Comfortable across academia and industry, he blends deep theoretical work with practical systems engineering, including shader fixes, network streaming, and multi-face video processing. An interesting detail: his projects combine low-level native/Unity optimization with high-level model estimation, enabling research prototypes to become interactive, deployable experiences.
10 years of coding experience
5 years of employment as a software developer
Doctor of Philosophy (PhD) Computer Science, Doctor of Philosophy (PhD) Computer Science at Warsaw University of TechnologyComputer Science, Computer Science at Universidad Politécnica de Madrid
Polish, English, Spanish
Github Skills (27)
opencv10
python10
net10
dotnet10
networking10
numpy10
asp-net10
3d-models10
gesture-recognition10
face-alignment10
hololens10
dotnet-core10
computer-vision10
dlib10
csharp10
Programming languages (4)
Github contributions (5)
MarekKowalski/LiveScan3D
Oct 2015 - Mar 2021
LiveScan3D is a system designed for real time 3D reconstruction using multiple Azure Kinect or Kinect v2 depth sensors simultaneously at real time speed.
Role in this project:
Full-stack Developer
Full-stack DeveloperContributions:1 release, 31 commits, 8 PRs in 5 years 5 months
Contributions summary:Marek primarily worked on expanding the functionality of the LiveScan3D system, focusing on both server-side and client-side components. They implemented network streaming capabilities for the merged point clouds, enabling data transmission and display on devices like HoloLens. Further contributions include client-side calibration data saving and loading, and the addition of a .ply file reader for the LiveScanPlayer application. The user also incorporated new features into the live view window.
reconstructionspeedsensorssimultaneouslykinect
MarekKowalski/FaceSwap
Jun 2016 - Feb 2018
3D face swapping implemented in Python
Role in this project:ML Engineer
ML EngineerContributions:14 commits, 10 pushes, 1 branch in 1 year 8 months
Contributions summary:Marek primarily contributed to the `faceswap` project by implementing and refining core functionalities related to 3D face modeling and swapping. They developed and modified key Python modules, including `models.py` and `utils.py`, to enhance the model parameter estimation, face keypoint detection, and texture mapping processes. The user also added features such as support for multiple faces in a single video frame and the ability to save the face-swapped results to a video file, improving the project's utility. Additionally, the user optimized performance by scaling down input images for faster face detection.
pythonswappingcomputer-vision3d-faceface
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.