Hi, I'm Youssef Elebiary

A comprehensive disaster response platform that leverages Multimodal AI to coordinate rescue operations. By fusing aerial surveillance (YOLOv11), Graph Neural Networks (GNN), and NLP, it enables dynamic routing on damaged roads and automated triage of distress signals.

• Dynamic routing via GNN safety prediction and millisecond-level C++ engine.
• Real-time survivor & hazard detection using dual-spectrum drone feeds.
• Automated distress message triage with multi-level priority classification.
• Interactive React website with live map visualization and instant route inference.

An intelligent web application that generates descriptive captions for uploaded images. Using deep learning techniques, it analyzes visual content and produces human-like descriptions, powered by a custom-trained model based on InceptionV3 and LSTM architectures.

• Upload any image and get an accurate description.
• Combines InceptionV3 for feature extraction with LSTM for sequence generation.
• Advanced text generation algorithm for coherent descriptions.
• User-friendly web interface for easy interaction.

A deep learning computer vision AI model used to detect cracks in concrete using ResNet18, developed with Python

• Classifying images according to the presence of cracks

TruthGuard

A web application that uses machine learning algorithms (XGBoost and LightGBM) to detect fake news. Users can input text and select the desired model (standard or lite) and the prediction of the model will be displayed.

• Predicting whether a given text is fake or real.
• Supports XGBoost and LightGBM prediction.
• Simple and responsive web interface.
• Automatic cleaning and preprocessing of the given text for accurate predictions.

NoMoreSpam

A web application that uses a machine learning algorithm (SVM) to classify if an email is spam or ham. Users can input the text content of the email and the model will display its prediction with the percentage of the email being spam.

• Predicting whether the given text is spam or ham.
• Displays the likelihood of the email being spam.
• Simple and responsive web interface.
• Automatic cleaning and preprocessing of the given text for accurate predictions.

RecipeHub

A platform for viewing cooking recipes with a simple UI, providing the user with the ability to search and filter recipes. Providing the ability to seek help from Semsema the builtin AI Agent that can answer questions.

• Giveing the ability to search for recipes with their name.
• Filtering recipes according to their type or ingredients or both.
• Chatting with Semsema to help in using the website.
• Saving favorite recipes.
• Simple and responsive web interface.

PhotoWizard

A GUI app used to apply different filters to images in various file formats developed with C++ and the Qt framework

• Applying multiple filters
• Live preview of modifications
• Removing all applied filters
• Undoing last applied filter

CSVsmith

A GUI app used to view and modify CSV files in a tabular form, developed with C++ and the Qt framework

• Viewing CSV files
• Modifying CSV files
• Inserting/Deleting rows and columns

Course Management System

A Course Management System developed using Microsoft SQL Server for the database management and C# for the user interface. Providing functionality for managing students, instructors, courses, and enrollments in an academic environment.

• Managing students and instructors
• Adding and assigning courses
• Recording and updating grades
• Viewing detailed reports using SQL queries
• User-friendly interface

PDFToolKit

A GUI app used to modify PDF files, developed with Python and customtkinter

• Splitting a PDF file at a given page number
• Merging two PDF files
• Extracting a page from a PDF file
• Inserting a PDF file at a specific position

A console-based app used to simulate cryptocurrency trading implementing the Merkle tree, developed with C++

• Reading data from CSV files
• Simulating moving through time

LowLevelMachine

A console-based app used to simulate how programs run at a low-level, developed with C++

• Supporting 13 different instructions
• Storing integers in 2's complement format
• Storing floating point numbers in 8 bits