I am an aspiring AI engineer and computer engineering graduate with hands-on experience in machine learning, deep learning, computer vision, and data analysis. I enjoy solving problems and have demonstrated this through participation in ICPC and securing 5th place in a national hackathon. I bring strong skills in networking, programming, and communication, with a solid foundation in AI, data science, and teamwork. I am eager to contribute innovative solutions and grow in dynamic tech environments. I have pursued hands-on training in AI and data science, completed internships and trainee programs with reputable organizations, and continually expand my toolset across Python, ML frameworks, cloud services, and data analytics. I thrive in collaborative settings and aim to deliver impactful, scalable solutions in fast-paced tech environments.

Youssef Salah Ahmed

I am an aspiring AI engineer and computer engineering graduate with hands-on experience in machine learning, deep learning, computer vision, and data analysis. I enjoy solving problems and have demonstrated this through participation in ICPC and securing 5th place in a national hackathon. I bring strong skills in networking, programming, and communication, with a solid foundation in AI, data science, and teamwork. I am eager to contribute innovative solutions and grow in dynamic tech environments. I have pursued hands-on training in AI and data science, completed internships and trainee programs with reputable organizations, and continually expand my toolset across Python, ML frameworks, cloud services, and data analytics. I thrive in collaborative settings and aim to deliver impactful, scalable solutions in fast-paced tech environments.

Available to hire

I am an aspiring AI engineer and computer engineering graduate with hands-on experience in machine learning, deep learning, computer vision, and data analysis. I enjoy solving problems and have demonstrated this through participation in ICPC and securing 5th place in a national hackathon. I bring strong skills in networking, programming, and communication, with a solid foundation in AI, data science, and teamwork. I am eager to contribute innovative solutions and grow in dynamic tech environments.

I have pursued hands-on training in AI and data science, completed internships and trainee programs with reputable organizations, and continually expand my toolset across Python, ML frameworks, cloud services, and data analytics. I thrive in collaborative settings and aim to deliver impactful, scalable solutions in fast-paced tech environments.

See more

Experience Level

Expert
Expert
Expert
Expert
Expert
Expert
Intermediate
Intermediate
Intermediate
See more

Language

English
Advanced

Work Experience

AI & Data Science Trainee at Digital Egypt Pioneers Program (MCIT+Microsoft)
October 1, 2024 - September 20, 2025
Participated in an intensive AI, data science, and machine learning training program (in collaboration with Microsoft). Gained hands-on experience in ML technologies, data preprocessing, model development, and evaluation. Worked on collaborative projects applying Python, scikit-learn, and TensorFlow; contributed to real-time data analysis and problem solving during the program.
Network Engineering Trainee at Benya Technologies
August 1, 2022 - September 20, 2025
Developed proficiency in network fundamentals, routing, switching technologies, security, and troubleshooting techniques.

Education

Bachelor in Computer Engineering – Major in Artificial Intelligence at Galala University
January 11, 2030 - January 1, 2025
Bachelor in Computer Engineering – Major in Artificial Intelligence at Galala University - Egypt
January 11, 2030 - January 1, 2025

Qualifications

AWS Academy Machine Learning Foundations
January 11, 2030 - January 1, 2025
Data Analysis course from IBM online
January 11, 2030 - January 1, 2024
Competitive Programming Training
January 1, 2023 - September 20, 2025
Machine Learning and Deep Learning (Udemy)
January 1, 2023 - September 20, 2025

Industry Experience

Software & Internet, Education, Professional Services, Computers & Electronics, Media & Entertainment
    Fraud-Detection-with-ANN
    Project Overview This project applies Deep Learning to detect fraudulent transactions in the Fraud Detection Dataset. Originally, the implementation used a 1D Convolutional Neural Network (CNN), but the model was later adapted into a feed-forward Artificial Neural Network (ANN), which is better suited for tabular data. The goal is to classify transactions as: 1 → Fraudulent 0 → Non-fraudulent ⚙️ Tech Stack Python 3.10+ TensorFlow / Keras (Deep Learning framework) scikit-learn (data preprocessing & metrics) NumPy & Pandas (data manipulation) Matplotlib (visualizations) 📂 Dataset File: fraudTrain.csv Contains 1,296,675 transactions and 11 numeric features. Target column: is_fraud Since fraud data is highly imbalanced, the dataset was balanced by downsampling non-fraud records to match the number of fraud records. 🔄 Data Preprocessing Removed non-numeric columns. Balanced dataset by downsampling majority class. Split into training (80%) and testing (20%) using train_test_split. Features were standardized using StandardScaler. Converted labels to NumPy arrays. 🧠 ANN Model Architecture The CNN was replaced with a fully connected ANN: Input Layer: 11 features ↓ Dense(128, ReLU) + BatchNormalization + Dropout(0.3) ↓ Dense(64, ReLU) + BatchNormalization + Dropout(0.3) ↓ Dense(32, ReLU) + Dropout(0.2) ↓ Dense(1, Sigmoid) → binary output (fraud / non-fraud) Optimizer: Adam (lr = 0.0001) Loss Function: Binary Crossentropy Metrics: Accuracy, Precision, Recall, AUC 📊 Training Epochs: 70 Batch size: 192 Validation: 20% split Training and validation curves were plotted to monitor convergence. ✅ Model Evaluation On the test set, the ANN achieved: Test loss: 0.373902 Accuracy: 0.854812 Precision: 0.947855 Recall: 0.750833 AUC: 0.900842 F1: 0.837918 (Values may vary slightly depending on random seed and balancing.) 📉 Results & Insights Balanced performance across Precision, Recall, and F1 indicates the ANN handles class imbalance well. AUC ~0.90 shows strong discriminatory power between fraud and non-fraud. Dropout layers helped mitigate overfitting.
    Generative Adversarial Networks for Improved Arabic Handwritten Characters
    Introduction This project addresses Arabic handwritten digit recognition using deep learning. A Generative Adversarial Network (GAN) is applied to generate synthetic data, and a CNN-BiLSTM classifier is trained on real and synthetic images. Dataset Training: 13,440 images (32×32 grayscale, CSV format). Testing: 3,360 images. Labels mapped to classes 0–9. Methodology Preprocessing: Reshape CSV → (32, 32, 1), normalize pixels to 0,1]. GAN: Generator creates synthetic digit images. Discriminator distinguishes real vs. fake. CNN-BiLSTM Classifier: CNN extracts spatial features. BiLSTM captures sequential dependencies. Softmax layer classifies into 10 digits. Inference: Function accepts an image file, preprocesses it, and outputs predicted digit + probability. Results GAN produced realistic synthetic digits. CNN-BiLSTM achieved high accuracy on test data (>90% typical). Single-image inference works reliably. Conclusion Combining GAN augmentation with a CNN-BiLSTM classifier improves Arabic digit recognition. Future work may include DCGAN, additional data augmentation, and deployment as a real OCR application.
    Intelligent PDF Question Answering & Summarization System
    Intelligent PDF Question Answering & Summarization System Using LSTM Embeddings and Google PaLM API 📌 Overview This project is an AI-powered multi-document understanding system that enables users to: Upload one or more PDF files. Ask questions and receive context-aware answers. Summarize PDF content efficiently. It combines: Custom LSTM embeddings for semantic text understanding. Google PaLM API (via LangChain) for natural language responses and summarization. Streamlit for a user-friendly interface. 🚀 Features 📥 Multiple PDF Support – Process and query several PDFs at once. 📑 PDF Summarization – Generate concise summaries of documents. 🧠 Custom LSTM Embedding Generation – Learns context from your data. 🔍 Context Matching Without FAISS – Retrieves relevant chunks via embeddings. 💬 Conversational Q&A – Uses Google PaLM for detailed, context-aware answers. ⚡ Fast & Lightweight – Works in real time after processing. 🛠️ Technologies Used Python Streamlit – UI framework TensorFlow / Keras – LSTM model building LangChain – Conversational & summarization chains Google Generative AI (PaLM) – Answer & summary generation PyPDF2 – PDF text extraction NumPy, scikit-learn – Data processing and training utilities ⚙️ How It Works Upload PDFs – One or more files via the sidebar. Text Extraction – Extract text from all PDFs. Text Chunking – Split into manageable, overlapping chunks. Embedding Generation – Train an LSTM model to generate embeddings. Question Answering – Retrieve the most relevant chunk and get an AI-generated answer. Summarization – AI condenses document content into a concise summary. ▶️ Usage Upload single or multiple PDFs. Click Submit & Process to train and save embeddings. Ask questions or request a summary.