At Hertel. Robotics, I led early-stage perception and autonomy system design for a UAV, contributing to its initial architecture and technical planning. I evaluated approaches for GPS-denied, windy, and maritime environments.
At Aalto University, I developed and evaluated computer-vision methods for 3D mapping, visual localization, and semantic understanding of urban environments. I implemented PyTorch-based perception models and integrated components into ROS-based systems, working with real-world data.
My PhD research also included a deep-learning approach to detect traffic-rule changes from real-world induction-loop measurements. I adapted a U-Net-based architecture for spatio-temporal data and evaluated it with simulated and real-world traffic data.
At Universitat Pompeu Fabra, I contributed to a learning-based visual relocalization system, integrating scene-coordinate prediction with geometric pose estimation. My work focuses on computer vision and robotics, including Structure from Motion, SLAM, and camera pose estimation.

