Mohamed is a Data Scientist providing expertise on the full project lifecycle of data solutions from analytic problem definition through data gathering, analysis, model development, reporting/visualization development, testing, and operational deployment. He applies deep learning and computer vision techniques in building models, performing statistical analysis and creating visualizations to provide scalable solutions on projects.
Identify opportunities for leveraging company data to drive business solutions.
Helps build and improve infrastructure, application and performance development and ensures tight security including data encryption, security groups, and environment scanning.
Closely collaborate with different functional teams to implement models and monitor outcomes.
MammoSUGGEST AI leverages the power of artificial intelligence (AI) to improve breast cancer detection in mammograms. It provides a second reader for radiologists to increase the confidence of diagnostic decisions, decrease recall rates, and prioritize readings.
MammoSUGGEST reads and automatically detects and classifies breast abnormalities in mammograms using state-of-the-art proprietary AI models. Worked as the main AI developer on the project - implementing solutions for breast cancer detection and classification from mammograms.
Actively participated in a six-member R&D team that worked on the computer vision library inside 360Imaging's system. Designed and translated research and technical requirements into tasks - working on new features, adding 2D active contour functionality, implementing 3D Geodesic Active Contours to help segment teeth from the volume, and refactoring the library modules to improve basic functions.
The Anomaly Detection solution uses a 120k+ endoscopy images dataset acquired from several sources for detection. It comes with features for image processing, dataset cleaning and preparation and uses unsupervised techniques to explore the images domains (One-Class SVM and isolation forest). Trained the GAN model to achieve 0.89 accuracy levels on the test set - using AWS virtual machine for model training.
