Mina Sorial

Lead the R&D team, driving innovation and development in computer vision and surveillance technologies.

Project: LiDAR Point Cloud Filtering for Autonomous Vehicles. Skills: Python, C++, Linux, ROS, CMake, Git, PCL, Docker • Developed a ROS1-based C++ pipeline to filter ground and noise from LiDAR point clouds while preserving key features like buildings and trees. • Implemented downsampling, vehicle point removal, RANSAC ground segmentation, and SOR/ROR noise filtering using PCL.

Project: Apple Detection System Skills: YOLOv8, PyTorch, OpenCV, Python • Developed a fruit detection pipeline using YOLOv8 in PyTorch to locate apple centers in orchard images under challenging lighting and occlusion conditions. • Evaluated traditional CV techniques (color/texture/shape analysis) vs. deep learning, demonstrating YOLOv8’s superior accuracy and robustness.

Balanced responsibilities effectively alongside full-time employment. • Introduced young minds to software engineering and artificial intelligence, inspiring their enthusiasm and curiosity • Gave interactive online coding and robotics sessions to students aged 6 to 18. • Developed an engaging curriculum integrating Python, AI, and robotics concepts to enhance student understanding and skills.

• AI Model Training: Contributed to the training of generative AI models to enhance their coding generation capabilities. • Code Evaluation: Assessed the quality of AI-generated code by writing detailed, human-readable summaries explaining the rationale behind evaluations. • Test Case Development: Developed robust test cases to ensure the accuracy and efficiency of code, verifying the performance and reliability of AI-generated solutions.

• Developing computer vision algorithms for object detection and tracking. • Actively contributed to a Proof of Concept(PoC) project focused on developing a perception system for an autonomous vehicle.

Project: Obstacle Detection System for Unmanned Surface Vehicle. • Developed and implemented a robust tracking-by-detection pipeline for camera data in USV environments, utilizing Yolo CNN for accurate object detection, and incorporating Kalman filter to enhance tracking performance. • Performed clustering and segmentation of obstacles in the 3D LIDAR point cloud. • Successfully tracked obstacles, estimating their position, velocity, and orientation in the 3D reference frame.

Project: Real-time Platooning in an unknown environment within VALET Project Skills: C++, Linux, ROS. • Implemented a range of longitudinal and lateral control laws to ensure stable and secure platoon formation. • Conducted comprehensive testing and evaluation of these control laws using C++ and ROS • Integrated platoon functionality into the existing autonomous fleet. • Validated control laws and platoon functionality through real-world testing, demonstrating reliability, performance, and safety.