Selected Work
A project developing new technology for autonomous trash collection on water, developed as part of the Mubadala trash-collection ASV competition. The platform brings together three parts: a collection robot (Burger), a reactive vision navigation system (SUSHI), and a development simulation (Toast).
A seabin-inspired collection mechanism built as a compact combat robot. Smaller and cheaper than full-scale systems, designed to carry far more trash to scale.
Led by Saif Alsaad (Mechanical), Laith Mohamed (Electrical), and Abdul Maajid Aga (Embedded).
The vision system for reactive, uninformed surface navigation: water-aware perception, object detection, depth estimation, multi-field path planning, and behaviour-based control.
The simulation environment built to develop SUSHI in software first, then validate it through real-world testing.
Led by Muhammad Abban.
E-puck robots pre-evolved in empty environments adapt Hebbian-modified MLP weights to novel obstacles on-the-fly in Webots, without retraining.
A multi-modal beehive monitoring system for honeybee health and hazard detection, combining predator detection, environmental sensing, and acoustic analysis. Currently a working abstract with a trained queen-less CRNN and bee/hornet detection, not yet deployed in a conditioned environment.
Predator detection with YOLO (bee vs hornet), environmental sensing (temperature, CO2, humidity, weight), and a CRNN sound module for queen-less detection, together forming an early hazard and colony-collapse warning system.
Related paper →MSc thesis project. VQ-VAE applied to PaSST embeddings of honeybee recordings to discover non-semantic acoustic sub-states without labels, building on BeeBetter's acoustic foundation.
Related paper →Developed a robot audition system for sound azimuth localization using binaural processing with a two-microphone array on an Arduino-based EMoRo robot platform.
Related paper →Interactive path planning system for a 7-DOF Franka Emika Panda robot in PyBullet. Implements RRT, A*, STOMP and polynomial interpolation from scratch with orientation-aware configurations, real-time obstacle avoidance, path smoothing, and live 3D trajectory visualisation.
Designed and implemented a 2-axis camera mechanism for real-time facial tracking with precise motor servoing through 360° servos on embedded hardware, using Google Coral TPU and TensorFlow Lite.
Facial recognition database using Redis for user verification with liveness detection, deployed across private residence security systems.
Multi-detection system for extracting vehicle colour and license plate information through YOLO-based detection and OCR plate number extraction.
Background
Laboratory Assistant
Heriot-Watt School of Mathematics & Computer Science · Dubai
Teaching Assistant
Heriot-Watt School of Engineering & Physical Sciences · Dubai
Embedded Systems & ML Engineer
Vee Tech · Cairo
Backend Intern
Vee Tech · Cairo
Outreach
Dubai Future Labs · 5-Day Intensive
5-day intensive on MoveIt, SLAM, mapping, and OpenCV in ROS2, organised by Dubai Future Labs and The Construct.
Heriot-Watt University Dubai · Volunteer
Led Arduino programming and obstacle-detection workshops for high-school visitors. Operated the Unitree GO1 quadruped during multiple outreach events.
Khalifa University · Abu Dhabi, UAE
Hardware demonstrations and interactive robotics competition at one of the world's premier robotics research conferences. Guided attendees through exhibits and hands-on demos.
ITS World Congress · Volunteer
Operated and demonstrated university robots to visitors, engaging the public with live robotics demonstrations.