Munich, Bavaria, Germany
🚀 About Me Robotics Software and Hardware Engineer specializing in automation and AI. With a Master's in Mechatronics from the Karlsruhe Institute of Technology (KIT), I focus on innovative solutions for imitation learning, dexterous manipulation, and multi-agent systems. My expertise spans ROS (1 & 2), Python, C++, and computer vision, with proven experience taking robotic concepts from simulation to real-world application. 👨💻 Professional Highlights My professional experience includes developing a ROS2 package for automatic camera calibration as a Scientific Research Assistant at KIT's Institute for Material Handling and Logistics (IFL). Previously, at Seamless Engineering, I programmed multi-agent systems (UARMs, TurtleBots) in ROS1 and developed Linux-based robotics software simulated in Gazebo. 🎓 Academic Highlights My Master's thesis at KIT focused on "Imitation Learning for Dexterous Manipulation (Shadow Hand + URe10) using Latent Space Representations for Probabilistic Movement Primitives (ProMP) training". Other key projects include a haptic feedback robot for neuromotor rehabilitation and an urban search and rescue vehicle. 💡 Core Competencies Robotics & Control: ROS 1 & 2, MoveIt, Motion Control, Kinematic Path Planning. AI & Machine Learning: Imitation Learning, Reinforcement Learning, PyTorch, LLM Integration. Computer Vision: OpenCV, Camera Calibration, Neural Networks, LVM. Programming & DevOps: Python, C++, C, Git/GitLab CI, Linux. Simulation & CAD: Gazebo, NVIDIA Isaac Sim, SolidWorks, 3D Printing. 🌍 Multilingual Fluent in Arabic, English (C2), and German (C1), I am an effective communicator in diverse, international teams. 🔍 Career Goals I am seeking challenging roles in automation and AI, particularly those focused on dexterous manipulation and multi-agent systems, where I can apply my skills to solve complex industrial problems. 💼 Let's Connect! I am open to connecting with professionals in robotics, AI, and mechatronics to exchange ideas and explore collaborations.
For my Master's thesis Imitation Learning for Dexterous Manipulation Utilizing Latent Space Representations in Dynamic Movement Primitives, I developed a complete pipeline for teaching a robot complex manipulation tasks from a single human demonstration. The project is divided into three core stages: First, we developed a robust preprocessing pipeline to handle raw robot data. Human demonstrations, recorded as high-dimensional 30D state vectors (6D arm pose + 24D hand joints), are systematically cleaned. This involves smoothing sensor noise, correcting for rotational discontinuities in the arm's orientation, and, most importantly, using dimensionality reduction techniques like PCA and UMAP to compress the complex 24D hand joint data into a compact and meaningful 3D latent representation. This stage transforms noisy, high-dimensional data into a clean, low-dimensional 9D format ready for learning. Next, we built a decoupled imitation learning model using Probabilistic Movement Primitives (ProMPs). Instead of a single monolithic model, we train two independent ProMPs: one for the 6D arm pose and another for the 3D latent hand synergy. This design prevents the arm’s movements from incorrectly influencing the hand’s grasp and vice-versa, enhancing generalization. Each ProMP learns a probabilistic distribution from the demonstration, capturing not just the average movement but also its inherent variability. Finally, we introduced human-in-the-loop in an intuitive interactive editing and generalization workflow using a "bookmarking" feature. A user can play back the robot's learned trajectory, pause at a key moment (e.g., the pouring location), and "bookmark" that state. By simply editing the coordinates of this bookmarked via-point, the user can define a new goal. The system then uses ProMP conditioning to automatically generate a new, smooth trajectory that passes through the new point while preserving the overall learned skill and achieving generalization.
Es freut mich mitzuteilen, dass ich derzeit Teil der Gruppe für künstliche Intelligenz in der Robotik (AIR Group) am Institut für Fördertechnik und Logistik des Karlsruher Instituts für Technologie bin.
Completed a 5-month Seamless Engineering internship focused on developing autonomous multi-agent robotics systems for warehouse logistics applications. Key Responsibilities & Achievements: • Developed and programmed multi-agent systems integrating UArm manipulators, TurtleBot mobile robots, and conveyor belt systems using ROS 1 and Python • Collaborated in a 6-member cross-functional team utilizing GitLab for version control and agile methodologies (Scrum) • Conducted comprehensive simulation and testing of robotic systems using Gazebo simulation environment • Performed project management tasks including requirements analysis, resource planning, and execution • Implemented Linux-based software development practices for robotics applications • Contributed to system architecture design and ROS network configuration for autonomous warehouse operations