Data Scientist and AI Developer in Cairo, Cairo Governorate, Egypt

A system that helps human annotators label image pixels for semantic segmentation. The system uses active learning to suggest a subset of the pixels to be labeled, yet arrives at comparable accuracy. I built and implemented the system using PyTorch and other helping libraries. It significantly reduces the effort required by a human annotator to label an image.

A system that matches colleagues with similar objectives and skills to facilitate collaboration. I built a system that uses word-embedding algorithms to find the semantic similarity of employees based on their profiles. Employees can easily use the system to find colleagues to help them with their tasks.

An automated pipeline, created with Python and Bonobo, runs chained data processing operations with multiple parameters and automatically produces analysis plots with Seaborn. The goal of the pipeline was to reverse engineer the parameters needed to replicate some results for the client.

A system that uses Generative Adversarial Networks to recover high-resolution images from low-resolution images and semantic labels used by the client. The system is implemented in PyTorch and uses the pre-trained model of the SuperResolution GAN architecture.

A deep learning pipeline for generating traffic scene images for training self-driving cars. The pipeline uses recent advances in graph convolutional neural networks and generative adversarial networks to control the number and type of objects in the traffic scene and the generated images' daytime and weather conditions.