I proposed the EventRec project and led it from concept to securing funding from the AIF/IGF, in close collaboration with colleagues.

The project focuses on perception for high-speed robotics (e.g., the O³dyn robot) using event cameras. It develops generic methods with a multi-modal stereo-event and RGB camera system for 3D detection of fast-moving objects, enabling mobile robots to navigate at high speeds.

Industry partners: KION, Jungheinrich, Framos and others.

Package for 2D segmentation of unseen objects and the main outcome of my PhD work.

The package segments arbitrary objects from only a few images, making it broadly applicable. I focused on two applications: bin picking and grasping for mobile robots, for which dedicated datasets were collected.

While the package itself targets segmentation, our centroid triplet loss (CTL) built on top of it achieved strong results on ArmBench. This line of work is especially relevant to logistics, where countless items must be segmented and classified, and parts have already been transferred to industry.

I'm involved in the BOP community which is organized by Meta, Google, CTU Prague, ENPC ParisTech, NVIDIA, Niantic, and others.

BOP is a collaborative benchmark for 6D pose estimation. I joined the organizers in 2024, contributing to the annotation tool in the BOP toolkit, dataset collection, and toolkit maintenance. The benchmark has recently expanded to unseen 6D pose estimation, closely aligned with my PhD research.

Developed a grasping pipeline integrating deep learning and ensemble methods, including dataset collection, analytical approaches, and a voting mechanism to boost performance.

DeBr is a radio protocol designed to efficiently flood information across dense, unstructured networks of radio nodes. It leverages constructive interference, where multiple radios transmit the same message simultaneously with nanosecond precision. I was responsible for implementing the entire protocol and developing its firmware using Contiki-NG OS and the TI SimpleLink SDK. This work was part of my then supervisor Aswin Venkatapathy’s PhD research (now ETH Zürich).

Sensor-Floor is a grid of 345 Texas Instruments SensorTags, each equipped with ten different sensors. The testbed is used to evaluate algorithms related to radio protocols (e.g., DeBr) and robot localization. The project was mainly developed by Aswin Venkatapathy.

I was responsible for programming the node firmware and developing a Linux command-line tool to control the system, including flashing the nodes and collecting data. Additionally, I contributed to the development of a localization system built on top of the Sensor-Floor platform.

The project aimed to develop a semi-autonomous mobile robot for landmine detection. My responsibilities included sensor fusion of laser and stereo cameras for point cloud acquisition, integrating RTK-GPS and IMU data for localization, as well as hardware setup and PCB design.

A collection of self-initiated robotics projects and competition entries during my early years in Egypt. These projects involved building robots from scratch, experimenting with ROS, sensor fusion, embedded control, and maze-solving algorithms.