Pinned Repositories
Ahmad-Sawalqeh
About me
AI-Generated-Blog-Generator
Developed a Streamlit app generates a blog post using AI models from Hugging Face. Users input a blog title, and the app creates an Introduction, Body, and Conclusion based on that title. It uses a text generation model to create the content and an image generation model to create the images.
Airbnb-Listing-EDA
Performed an exploratory data analysis on the Airbnb listings dataset for Amsterdam, revealing that the highest average prices were recorded in 2023, while the lowest were in 2014. Prices peaked in April and were at their lowest in March, whereas availability was highest in June and lowest in September.
Arabic_Abstractive_Summarization
Fine-tuning the mT5 transformer model for abstractive text summarization in Arabic, aiming to produce concise and meaningful summaries. The model's performance was evaluated with ROUGE metrics, showing preliminary results with ROUGE-1: 0.10, ROUGE-2: 0.02, and ROUGE-L: 0.08. This work is part of the Competition Military Technical College.
artificial_neural_networks
This repository is your comprehensive guide to building Artificial Neural Networks (ANN) using various frameworks in Python. Explore building ANNs from scratch using NumPy, implementing ANNs with TensorFlow, and creating ANNs with PyTorch.
auto-annotate
A simple tool for automatic image annotation using Roboflow API
automated-scrape-events
building an automated tool to scrape events content from local websites and have them ready to be inserted in a database which can eventually be used in our mobile app (examples: Visit Qatar, ILoveQatar, Qatar Calendar, QF Website, QU Website, etc.). We plan to use the NER (Named Entity Recognition) technique and/or a Large Language Model (LLM).
automated_eda
Developed an AutoEDA platform that streamlines the complete lifecycle of exploratory data analysis, including data acquisition, preprocessing and visualization. Designed for users with minimal data science expertise.
deep_neural_networks
This repository is your comprehensive guide to building and working with deep neural networks (DNNs) in Python. Explore a wide range of topics, from generating and preprocessing data to training and inference, as well as advanced techniques such as embedding layers, convolutional neural networks (CNNs), long short-term memory networks (LSTMs).
naive-bayes-LSTM-for-sentiment-analysis-NLP-widebot
Developed a sentiment analysis to classify text into positive, negative, or neutral. Utilized NLTK for preprocessing and fine-tuned the TF-IDF vectorizer and Naive Bayes classifier with RandomizedSearchCV, achieving 93% validation accuracy and 91% testing accuracy. Additionally, implemented LSTM model with 92% validation and testing accuracies.
elsayedelmandoh's Repositories
elsayedelmandoh/naive-bayes-LSTM-for-sentiment-analysis-NLP-widebot
Developed a sentiment analysis to classify text into positive, negative, or neutral. Utilized NLTK for preprocessing and fine-tuned the TF-IDF vectorizer and Naive Bayes classifier with RandomizedSearchCV, achieving 93% validation accuracy and 91% testing accuracy. Additionally, implemented LSTM model with 92% validation and testing accuracies.
elsayedelmandoh/AI-Generated-Blog-Generator
Developed a Streamlit app generates a blog post using AI models from Hugging Face. Users input a blog title, and the app creates an Introduction, Body, and Conclusion based on that title. It uses a text generation model to create the content and an image generation model to create the images.
elsayedelmandoh/Airbnb-Listing-EDA
Performed an exploratory data analysis on the Airbnb listings dataset for Amsterdam, revealing that the highest average prices were recorded in 2023, while the lowest were in 2014. Prices peaked in April and were at their lowest in March, whereas availability was highest in June and lowest in September.
elsayedelmandoh/Arabic_Abstractive_Summarization
Fine-tuning the mT5 transformer model for abstractive text summarization in Arabic, aiming to produce concise and meaningful summaries. The model's performance was evaluated with ROUGE metrics, showing preliminary results with ROUGE-1: 0.10, ROUGE-2: 0.02, and ROUGE-L: 0.08. This work is part of the Competition Military Technical College.
elsayedelmandoh/automated_eda
Developed an AutoEDA platform that streamlines the complete lifecycle of exploratory data analysis, including data acquisition, preprocessing and visualization. Designed for users with minimal data science expertise.
elsayedelmandoh/automated_ml
Developed an AutoML platform that optimizes the entire lifecycle of machine learning models for classification and regression tasks, from data acquisition to model saving, achieving 100% usability with zero-code features, designed for users with minimal data science expertise.
elsayedelmandoh/CARD_MATCH_CHALLENGE
Developed a Card Match game using Tkinter, where players swap cards to match their own hand or the main deck. The game ends when either all of the player's cards or the main deck's cards are identical, and the state is updated dynamically with each turn.
elsayedelmandoh/code-evaluator
Developed a Streamlit app to evaluate code solutions. Users can input task descriptions and code solutions to receive detailed feedback on task alignment, modularity, efficiency, and adherence to key concepts, as well as verify if the code is AI-generated.
elsayedelmandoh/customer_segmentation
Developed a K-means clustering model for customer segmentation and personalization in an e-commerce context. Utilized PCA for dimensionality reduction and WCSS for cluster validation, identifying three customer groups with similar preferences and behaviors.
elsayedelmandoh/dashboard_ElectroPI
Developed a Streamlit dashboards for visualizing and analyzing data within an organization's system. The dashboards provide interactive insights into user activity, subscriptions, course completion, capstone evaluations, coupon usage, and employment grant status.
elsayedelmandoh/detect-terrorists
Developed a YOLOv8x model with ultralytics to identify potential terrorists by classifying individuals as civilians or military and detecting weapons. Utilized AutoDistill for annotation and autolabeling, and defined ontology to enhance model training. Achieved a final accuracy of 0.738 mAP50 and 0.637 mAP50-95. Competition Air Defense College
elsayedelmandoh/Devices-Price-Classification-System-using-Python
Developed a Devices Price Classification System utilizing a pipeline that incorporates StandardScaler, SelectFromModel, and RandomForestClassifier, achieving an accuracy of approximately 93% in predicting device prices.
elsayedelmandoh/elsayedelmandoh
elsayedelmandoh/evaluating-language-models-for-harmful-prompt-detection
Evaluated and compared language models for harmful prompt detection using LangChain and prompt engineering techniques. Implemented and tested Google Gemini 1.5 Flash, achieving 62% accuracy, and Anthropic Claude 3 Sonnet, achieving 52% accuracy.
elsayedelmandoh/GemiLla
GemiLla is a Streamlit-based project that allows two conversational agents, Gemini and LLaMA, to engage in interactive discussions. It demonstrates the integration of Google's Gemini and Meta's LLAMA-3, providing an engaging and flexible conversation interface.
elsayedelmandoh/generative-ai-for-beginners
18 Lessons, Get Started Building with Generative AI 🔗 https://microsoft.github.io/generative-ai-for-beginners/
elsayedelmandoh/LSI-LaBSE-semantic-search-widebot
Developed a semantic search pipeline to extract relevant keywords from English movie plots. Utilized SpaCy for preprocessing and Gensim for loading corpora, TF-IDF, and LSI models, achieving a Mean Average Precision (MAP) of 87%. Additionally, a pretrained model (LaBSE) was used for comparison.
elsayedelmandoh/market_segmentation_neural_networks
Developed a neural network with 4 Dense layers for market segmentation using customer data. The model achieved a mean absolute error of 2.8181 on training data and 2.6434 on testing data. This model helps businesses tailor marketing strategies to specific customer segments.
elsayedelmandoh/Medical-Image-Analysis-with-VGG16-ResNet-CNN
Developed multi-models to analyze X-ray images for detecting diseases or abnormalities. Utilized VGG16 with a training accuracy of 74% and a test accuracy of 77%. Implemented a ResNet model, achieving a training accuracy of 93% but a test accuracy of 62%. Additionally, trained a CNN model with a training accuracy of 74% and a test accuracy 62%
elsayedelmandoh/OCR_CNN
Developed a CNN model using TensorFlow to recognize alphanumeric characters with 98% accuracy on both training and testing datasets.
elsayedelmandoh/persoal_assistant_GPT-3
Developed Personal Assistant using OpenAI GPT-3 integrates AI-driven text and speech interactions. This tool offers intelligent responses through both spoken and written inputs.
elsayedelmandoh/problem-solving
Problem Solving using Python
elsayedelmandoh/recognition-id-eru
Developed a Streamlit app allows users to upload an image, processes it using Optical Character Recognition (OCR) with Tesseract to extract details like the name, university, faculty, and ID from the image and displays them in a table format
elsayedelmandoh/self-driving-car
Developed a small car controlled via a Wi-Fi application using a NodeMCU ESP8266. The system includes an L298N Motor Driver, gear motors, robot wheels, a battery holder, jumper wires, and Li-ion batteries.
elsayedelmandoh/Sentiment-Analysis-on-Social-Media-Posts-with-LSTM
Developed sentiment analysis to classify social media posts into positive or negative. Utilized NLTK for text preprocessing, TF-IDF for feature extraction, and LSTM model for classification, achieving 99% accuracy on the training data and 89% accuracy on the test data.
elsayedelmandoh/sentiment_analysis_NLP
Developed a sentiment analysis to classify text as positive, negative, or neutral. Utilized NLTK for preprocessing and fine-tuned the TF-IDF vectorizer and Naive Bayes classifier with RandomizedSearchCV, achieving 93% validation accuracy and 91% testing accuracy. Additionally, implemented an LSTM model achieving 92% validation and testing.
elsayedelmandoh/skin_cancer_mnist
Developed a skin cancer detection model using the VGG16 architecture and the MNIST dataset of skin cancer images. The project involved data preparation, augmentation, model building, and training. Achieved a validation accuracy of ~78%, demonstrating the model's potential in early skin cancer detection.
elsayedelmandoh/tic_tac_toc
Developed a Tic Tac Toe game where players input their moves on a 3x3 board. The objective is to align three symbols in a row, either horizontally, vertically, or diagonally.
elsayedelmandoh/twitter_disaster_classifier
Developed a sentiment analysis pipeline to classify disaster tweets. Utilized NLTK for preprocessing, and fine-tuned the TF-IDF vectorizer and Naive Bayes classifier with RandomizedSearchCV, achieving 85% accuracy. Additionally, trained LSTM model, achieving an accuracy of 85%.
elsayedelmandoh/Ubuntu-Automated-Customer-Service
Developed a Transformer-based model and fine-tuned GPT-2 for building an automated customer service chatbot using the Ubuntu Dialogue Corpus. The Transformer notebook covers data preprocessing, model architecture design, and training, while the GPT-2 notebook focuses on data preparation, fine-tuning, and saving the enhanced model.