Mehran Pashaei

* Developed a multi-task PyTorch framework for atrial fibrillation detection, diagnostic classification, signal quality assessment and ECG delineation. * Benchmarked CNN, CNN-Transformer and bidirectional Mamba encoders for performance and computational efficiency. * Evaluated self-supervised and semi-supervised learning to reduce reliance on labeled ECG data. * Tested robustness and generalization through controlled experiments and external validation.

* Built a Python/C++ ROS 2 node to correct GPS-based timing offsets across radar, LiDAR and camera streams, supporting reliable sensor fusion. * Migrated an autonomous vehicle software stack from ROS 1 to ROS 2, improving maintainability. * Automated builds and validation with GitLab CI/CD, reducing manual integration work. * Developed a LoRa communication bridge between distributed ROS environments and validated the sensor pipeline on a physical vehicle.

* Developed ML models to detect, classify and localize power-grid faults using real and simulated time-series data. * Built reproducible Python/PyTorch pipelines for preprocessing, labeling and training, enabling systematic model evaluation. * Stress-tested models against missing measurements, lower sampling rates and data loss to assess robustness. * Co-authored three peer-reviewed publications on ML for power-system protection, including real-grid disturbance analysis.

Funded by TÜBİTAK under the 2209-A University Students Research Projects Support Program. * Built an autonomous indoor robot with 2D LiDAR, wheel encoders and ROS control on Raspberry Pi. * Implemented Python ROS nodes and integrated Cartographer SLAM and the ROS Navigation Stack for mapping and obstacle-aware navigation. * Validated navigation in Gazebo and RViz before physical testing, supporting faster hardware iteration.

* Developed C firmware for microcontroller-based industrial automation systems. * Designed PCB layouts in Altium Designer and supported hardware testing and validation.

Specialization: Medical Image and Data Processing. Relevant coursework: Machine Learning for Engineers, Pattern Recognition, Pattern Analysis, Biomedical Signal Analysis, AI in Medical Robotics, Optimization for Engineers, Speech and Language Processing. Master's thesis: Multi-Task Learning for Electrocardiogram (ECG) Analysis with Self- and Semi-Supervised Representation Learning; Pattern Recognition Lab, in collaboration with MedTec & Science GmbH.

Bachelor's thesis: Development of an Autonomous Mobile Robot (TÜBİTAK 2209-A funded research project), supervised by Prof. Dr. Musa Alcı. Focus: robotics, ROS, SLAM, signal processing and electronics.