Experience

Work history, education, and publications.

Work History

Samsung Electronics

March 2023 - Present

Ported Cyberpunk 2077 to the Samsung Odyssey 3D glasses-free display with CD PROJEKT RED (one of 5 engineers): adapted HLSL shaders in REDengine for stereoscopic rendering and extended the Odyssey 3D SDK (C++) with a runtime 3D-toggle API; announced at gamescom 2026
Sole developer of a generative virtual try-on PoC: person + garment photos to a consistent 360° orbital video, combining Qwen-Image-Edit with Wan2.2 VACE video generation in ComfyUI, with an OpenPose control-video generator and an automatic geometric-consistency gate
Built a real-time 3D Gaussian Splatting viewer with stereoscopic rendering; moved a GPU matrix-inversion bottleneck to the CPU, raising frame rate from 25–30 to 40–45 FPS on RTX 4070
Trained and tuned Scaffold-GS on a single 12 GB GPU, balancing PSNR/SSIM/LPIPS against VRAM; built iOS (ARKit) and Android (ARCore) AR-guided capture apps for SfM + 3DGS datasets
R&D PoCs on facial landmark detection, background segmentation and Structure from Motion; literature review to working implementation
Stabilized a legacy sign language avatar system: introduced integration tests, raised static analysis (SAM/SVACE) scores from 60–70% to 85%; designed CI/CD in GitHub Actions

C++PythonPyTorchHLSLSwiftComfyUIDockerCMakeGitHub Actions

Mnemosis

July 2021 - February 2023

Co-developed MVP software for ski jumping biomechanical analysis in a 4-person Scrum team; used by 3 coaches of the Polish national ski jumping team
Perspective correction algorithm for spatial measurement of a moving athlete: projection error from >200 mm to <0.1 mm at 7.2 ms/frame (basis of my Master's thesis)
3D/4D skeletal fitting from multi-camera point clouds (100+ synchronized 4K cameras at 60 FPS): RMS mesh-to-point-cloud optimization with warm start, positional error <5 mm
Distributed multiprocessing pipeline for point cloud computation across local nodes; Autodesk Maya plugin and Unreal Metahuman workflow integration; Jenkins CI/CD

C++PythonOpenCVOpen3DEigenQt/QMLDockerJenkins

Smarttech3D Metrology

August 2020 - June 2021

Proof of concept of the company's new product based on Arduino
Implementation of algorithms and solutions for image processing or 3D scanning in C++
Working with LiDAR and photogrammetry
Prototyping solutions in Python

C++PythonOpenCVEigenArduinoLiDAR

Warsaw University of Technology

2021 - 2022

Warsaw University of Technology

2017 - 2021

DeepVID: deep-learning accelerated variational image decomposition model tailored to fringe pattern filtration

Maria Cywińska, Konstanty Szumigaj, Michał Kołodziej, Krzysztof Patorski, Vicente Mico, Shijie Feng, Chao Zuo and Maciej Trusiak

Journal of Optics, Vol. 25, No. 4 (2023) · [Article] · [PDF]

OpenCV · PyTorch · Structure from Motion · Gaussian Splatting · NeRF · 3D Scanning · Photogrammetry · LiDAR · ComfyUI · Stereoscopic Rendering

C++20 · Python · Swift · HLSL · Git · CMake

Algorithms and Data Structures · Algebra and Geometry · Optimization Problems

Docker · GitHub Actions · Jenkins · Static Analysis · Scrum · Code Review · Linux · Windows

Company logos are trademarks of their respective owners and are used here solely to indicate employment history.