Developed and maintained ROS-based software stack for unmanned aerial vehicles (UAVs). Designed flight logic and system architecture using ROS, C++ and Python. Integrated and configured flight controllers, onboard sensors, and vision pipelines. Optimized and deployed machine learning models for computer vision on embedded platforms (Rockchip, NPU). Performed end-to-end drone integration, including hardware assembly, calibration, and flight testing.

Tram Rail Segmentation & Defect Detection Technologies: Python, YOLO (v8/v11), PyTorch, OpenCV, SAHI, CVAT, Docker Key Contributions: Built a custom dataset of 10,000+ images from scratch (collection, annotation, augmentation). Developed a YOLO-based segmentation model for real-time rail detection, upgraded iteratively with new YOLO releases. Designed a two-stage defect detection system: Stage 1: Traditional CV methods for rapid defect localization. Stage 2: Fine-grained classification using a trained YOLO model. Researched and integrated SAHI framework to improve small-defect detection accuracy. Deployed models on trams for field testing, with iterative refinements based on performance feedback. Assessing the condition of drivers Technologies: Python, YOLOv11, ResNet, Wav2Vec 2.0, OpenCV Key contributions: Developed and trained computer vision models to detect driver distractions (phone usage, smoking) and emotional states (e.g., fatigue, stress) Integrated models into a real-time monitoring system to analyze tram driver behavior, triggering alerts for unsafe actions (e.g., inattentiveness, rule violations). Designed a notification module to provide actionable feedback to drivers and supervisors, enhancing adherence to safety protocols and reducing risks. Collaborated with cross-functional teams to deploy the system on edge devices, ensuring low-latency performance and reliability.

- Published 2 first-author papers and 1 co-authored paper on robotics and computer vision topics - Developed Gazebo simulation models for Alvus robot and compared ROS-based visual SLAM methods - Designed and constructed omnidirectional chassis for modular robotic platform Artem Apurin, Bulat Abbyasov, Liaisan Safarova. et al. (2023). LIRS-ArtBul: design, modeling and construction of an omnidirectional chassis for a modular multipurpose robotic platform. https://link.springer.com/chapter/10.1007/978-3-031-23609-9_7 Safarova, L. et al. (2023). Comparison of Monocular ROS-Based Visual SLAM Methods. https://link.springer.com/chapter/10.1007/978-3-031-23609-9_8 Safarova, L. et al. (2022). Alvus Modeling in Gazebo. https://kpfu.ru/staff_files/F1291110874/2022_Alvus_Modeling_in_Gazebo.pdf

Developed new functionality using Java and Spring Boot Wrote unit tests and debugged code Collaborated with team on internal projects Utilized Spring, Docker, and Git

Implemented new features with Java and Spring Boot Performed testing and bug fixing Participated in team-based development of internal projects Worked with Spring, Docker, and Git