Hi there, I'm Raafat Nagy, an AI Engineer focused on building intelligent systems across Computer Vision, Machine Learning, Deep Learning, Time-Series Forecasting, and NLP / RAG.
This repository serves as a central hub for my AI projects in these domains, covering both model development and end-to-end system design.
- Computer Vision Projects
- NLP and RAG Projects
- Deep Learning Projects
- Machine Learning Projects
- Time Series and Forecasting Projects
- Future Work
-
VisionSeek AI - Semantic Image Retrieval for Custom Datasets
- Built an end-to-end semantic image retrieval system supporting text-based, image-based, and hybrid search.
- Used Google SigLIP multimodal embeddings and FAISS vector search to retrieve visually relevant images using cosine similarity.
- Implemented dataset upload and validation, embedding generation, FAISS index building, and automatic index reloading.
- Developed REST APIs with FastAPI and a responsive React frontend for dataset management, search, result visualization, and similarity scoring.
- GitHub Repository - Demo Video
-
Background Removal Studio
- Built an AI-powered web application for high-quality background removal using FastAPI, PyTorch, and BiRefNet.
- Features transparent PNG generation, background replacement, GPU acceleration, Docker support, and a modern responsive UI.
- GitHub Repo - Demo Video
-
Vehicle Detection, Tracking, Counting and Speed Estimation
- Developed a real-time traffic monitoring system using YOLO and ByteTrack.
- Implemented speed estimation with perspective transformation for accurate measurements.
- Supported configurable counting zones and custom YOLO models.
- GitHub Repo - Demo Video
-
YOLO Object Detection App
- Built a real-time object detection web app with FastAPI backend and JavaScript frontend.
- Features drag & drop uploads, multiple model options, smart video streaming, and dark mode UI.
- Integrated advanced CV models with smooth asynchronous processing.
- GitHub Repo - Demo Video
-
Smart Face Attendance System
- Created a real-time face recognition attendance system using OpenCV and face_recognition.
- Automated attendance logging with webcam detection and CSV export.
- Optional API integration for backend synchronization.
- GitHub Repo
-
Student Entry and Exit Tracking
- Developed a system for tracking and counting students entering/exiting halls using YOLO and OpenCV.
- Used Shapely for zone-based direction detection.
- Enabled CSV logging and API reporting.
- GitHub Repo
-
Facial Landmark and Drowsiness Detection
- Implemented real-time facial landmark detection and drowsiness monitoring using the EAR method.
- Built with OpenCV and dlib for accurate fatigue alerts.
- Enhanced driver and workplace safety.
- GitHub Repo - Demo Video
-
Object Detection Telegram Bot
- Developed asynchronous Telegram bot for real-time object detection on user-submitted images.
- Automated annotation and detailed detection summaries.
- Built with OpenCV and python-telegram-bot for efficient performance.
- GitHub Repo - Demo Video
-
Chat with Any PDF using End-to-End RAG Assistant
- Built a RAG-based system to chat with PDF documents using natural language.
- Used LangChain, FAISS, HuggingFace embeddings, and Groq LLM for retrieval and generation.
- Developed a FastAPI backend with a simple web interface for document upload and querying.
- GitHub Repo - Demo Video
All projects are included in Deep Learning Projects repository:
-
Brain Tumor MRI Classification
- Developed a model using TensorFlow and ResNet50V2 for brain tumor detection from MRI images.
- Improved performance via data augmentation and transfer learning.
- Evaluated with confusion matrix and classification reports.
-
Oral Diseases Classification
- Built a multi-class classification model to identify six oral diseases using TensorFlow and ResNet50V2.
- Applied preprocessing, augmentation, and fine-tuned pre-trained layers.
- Assessed accuracy with detailed reports and confusion matrices.
-
Plant Disease Detection
- Designed a CNN to classify 38 plant disease categories with TensorFlow/Keras.
- Used batch normalization and dropout for better generalization.
- Achieved high validation accuracy through image augmentation.
-
MNIST Handwritten Digit Classification
- Created CNN for 10-class digit classification on grayscale images using TensorFlow.
- Applied data augmentation, dropout, early stopping, and learning rate scheduling.
- Validated with accuracy metrics and prediction visualizations.
-
Autoencoder Projects on MNIST
- Developed convolutional, simple, and denoising autoencoders for image compression and noise removal.
- Used convolution, max-pooling, and upsampling layers in encoder-decoder architecture.
- Evaluated reconstruction quality by comparing original and reconstructed images.
-
Machine Learning From Scratch
- Developed fundamental ML algorithms (Linear Regression, Logistic Regression, SVM, Decision Trees, KNN, Clustering, PCA) from scratch in Python.
- Emphasized mathematical understanding and clean, well-documented code for educational use.
- GitHub Repo
-
Diabetes Prediction Project
- Built ML models for diabetes prediction using patient health data with Python and scikit-learn.
- Conducted data exploration, visualization, feature engineering, and hyperparameter tuning.
- Deployed an interactive Streamlit app for real-time risk prediction.
- GitHub Repo - Streamlit App
-
Iris Flower Species Prediction
- Created an SVM model to classify Iris species based on sepal and petal measurements.
- Performed data preprocessing and exploratory analysis.
- Developed a Streamlit web app for user-friendly species prediction with dynamic visualization.
- GitHub Repo - Streamlit App
- Bitcoin Price Forecasting System
- Built an end-to-end time-series forecasting system for Bitcoin price prediction.
- Implemented multiple models including Prophet, ARIMA, and Auto-ARIMA with full evaluation pipeline.
- Developed an interactive Streamlit dashboard for forecasting, visualization, and uncertainty analysis.
- Handled financial time-series challenges such as volatility, missing data, and backtesting.
- Structured with a modular architecture (data processing, modeling, and UI layers).
- GitHub Repo - Streamlit App
This repository will continue to be updated with new projects and research work in the areas of:
- Machine Learning
- Deep Learning
- Computer Vision
- AI Systems Integration