At Aumovio, I build C++ components for visual SLAM and camera calibration, including bundle adjustment, state marginalisation and pose graph optimisation. I also developed a pose prediction module that fuses IMU, wheel encoder and GNSS data using vehicle motion models and Kalman filtering.
I'm researching online self-calibration of surround-view cameras and testing it in CARLA against ground truth. I also worked with my mentor on the research and architecture of a new calibration product, and ported and optimised C++ code for a real-time embedded target.
For my BSc thesis, I created a calibration and evaluation framework for automotive camera systems, achieving 0.206–0.350 px RMSE on KITTI and proprietary Aumovio rigs. At ELTE, I teach weekly Haskell practicals, and in my current MSc project I’m extending an RGB-D reconstruction framework with loop closure.

