Mariam Elsherbeny

I built a Digital Twin for Heart & Lung Dynamics, an interactive platform for patient-specific simulation that combines AI segmentation and forecasting models with FastAPI and real-time 3D visualization. As team lead and developer, I designed the database schema for patient data, simulation parameters, and model outputs; the project earned an A+ and placed second nationwide in Abdaa 13.

I've also built a real-time multimodal AI assistant using face, gesture, speech, and OCR recognition. I implemented DeepFace/VGG-Face authentication and MediaPipe with Random Forest gesture recognition, reaching 93% accuracy.

My NLP work includes an Arabic-to-SQL system fine-tuned from DeepSeek-Coder-6.7B with QLoRA and 4-bit quantization, as well as a RAG chatbot built with LLaMA-2, FAISS, and Gradio.

I'm completing a B.Sc. in Artificial Intelligence & Data Science at Egypt-Japan University of Science and Technology, with hands-on training in AI, machine learning, Python full-stack development, and SQL.