Mostafa Abbas Saad

About Me

AI Engineer specialized in Computer Vision, NLP, and Deep Learning

I have hands-on experience in designing and deploying end-to-end AI systems. I excel in building scalable, production-ready solutions using FastAPI, Docker, and MLflow.

My expertise includes Large Language Models (LLMs), RAG pipelines, and agent-based architectures, delivering high-impact AI solutions for real-world challenges.

Technical Skills

Machine Learning & Deep Learning

PyTorch, TensorFlow, Scikit-learn, CNNs, RNNs, LSTM, Transformers.

MLOps & Deployment

Docker, Docker Compose, FastAPI, MLflow, GitHub Actions (CI/CD), REST APIs.

LLMs & Generative AI

RAG, LangChain, LangGraph, FAISS, Prompt Engineering.

Computer Vision

YOLOv8, Object Detection, Image Classification, Segmentation.

NLP

Transformers, Embeddings, Text Processing, spaCy, Sentiment Analysis.

Programming

Python (OOP), Data Structures & Algorithms.

Featured Projects

End-to-End AI Predictive Maintenance System

Designed and deployed a full AI pipeline for Remaining Useful Life (RUL) prediction using LSTM (PyTorch), achieving 44% MAE reduction. Built a RAG-based diagnostic assistant with LangGraph and FAISS for context-aware insights. Developed REST APIs using FastAPI for real-time inference and integrated MLflow for experiment tracking and monitoring. Containerized the system with Docker Compose for scalable deployment.

PyTorch FastAPI Docker LangGraph

YOLOv8 Vehicle & License Plate Detection System

Developed a real-time object detection system using YOLOv8 for vehicles and license plates. Trained on a custom COCO-format dataset and optimized for real-world traffic scenarios. Achieved high detection accuracy under various lighting and occlusion conditions.

YOLOv8 Computer Vision

MRI Brain Tumor Segmentation

Built a U-Net-based deep learning model using PyTorch for medical image segmentation. Performed data preprocessing, augmentation, and fine-tuning to improve model generalization. Achieved accurate segmentation with IoU and Dice Score evaluation metrics, highlighting tumor regions reliably.

PyTorch U-Net Medical Imaging

Blood Cell Classification using Computer Vision

Developed a deep learning model using ResNet, EfficientNet, and VGG16 for blood cell classification. Implemented preprocessing, augmentation, and fine-tuning to maximize accuracy. Integrated the model into a Qt5-based GUI for interactive use and testing.

ResNet EfficientNet Qt5

Arabic-English Machine Translation System

Developed a Transformer-based neural machine translation model for bilingual translation. Implemented attention mechanisms for improved translation accuracy on Arabic-English datasets.

Transformers NLP

Ninja AI-LLM Chatbot

Built an LLM-based chatbot assistant to guide students in exploring computer science domains. Designed conversational flows and integrated knowledge-based responses for precise guidance.

LLMs Chatbot