Beshoy Gamal

Driven AI Developer and Computer Science student with strong practical and academic experience in machine learning, computer vision, and mobile app development, complemented by hands-on projects in healthcare, accessibility, and real-time AI solutions. Passionate about leveraging AI to solve real-world problems and create impactful applications. Seeking a junior AI developer or internship opportunity with an innovative company to expand expertise and contribute to forward-thinking projects.

Generative AI Trainee at Digital Egypt Pioneers Initiative – DEPI (2025-07 – Present)

Participating in an intensive training program focused on Generative AI, Large Language Models (LLMs), and applied machine learning techniques. Gaining hands-on experience with prompt engineering, fine-tuning, and deployment of AI models, as well as practical applications across NLP, computer vision, and multimodal systems.

• Enhanced Security Measures: Increased credit card fraud detection accuracy by 20%, improving transaction safety and reducing false positives by 35% in real-world financial data.
• Optimized Training Efficiency: Built a proprietary DNN training framework using NumPy, reducing model training time by 15% through optimized matrix operations and vectorization.
• Image Classification Success: Achieved a 10% higher accuracy on Kaggle's Cats vs Dogs dataset using a CNN with data augmentation and dropout regularization.
• Model Performance Improvement: Boosted Fashion MNIST classification accuracy to 87% through custom optimization, L2 regularization, and dropout.
• Pipeline Automation: Automated preprocessing and evaluation pipelines, cutting manual data preparation time by 40% and ensuring consistent model reproducibility.
• Model Deployment Optimization: Deployed trained models using lightweight Flask APIs, reducing inference latency by 25% and improving real-time performance.
• Data Quality Enhancement: Implemented advanced data cleaning and feature engineering workflows, increasing overall dataset consistency by 30%.
• Continuous Learning & Research: Conducted experiments with multiple optimizers (SGD, Adam, Momentum), improving convergence speed by 18% and model stability across training runs.

Bachelor in Computer Science – Fayoum University (2022-09)