Hi, I'm Arwa Elsawy
Junior Data Scientist · Software Engineer
I specialize in turning complex data into actionable insights and production-ready solutions. My experience spans predictive modeling, recommendation systems, web scraping, and deploying AI applications with Python, Odoo, Flask, and Hugging Face. I am passionate about solving real-world problems by bridging the gap between data science and practical business impact.
Skills: Python · SQL · Flask · TensorFlow · scikit-learn · Odoo · AWS · Docker
Technical Skills
- Programming: Python, JavaScript, C++, C#, SQL.
- Frameworks & Libraries: Flask, React.js, Node.js, TensorFlow, Keras, scikit-learn, Pandas, OpenCV, BeautifulSoup.
- Databases: PostgreSQL, MySQL, MongoDB, SQLite.
- Data & Analytics: ETL, data wrangling, feature engineering, predictive modeling, time-series forecasting, data visualization.
- Cloud & DevOps: AWS (EC2, S3), Docker.
- Tools & Platforms: Odoo ERP, Power BI, Selenium, Git, VS Code, PyCharm, Hugging Face.
Recent Projects
AI-powered Inventory Forecasting and Dynamic Slotting Optmization integrated within Odoo Framework (Graduation Project)
Problem
- Traditional inventory management in Odoo lacked accurate demand forecasting and efficient slotting.- This led to frequent stockouts and increased picking time.Approach
- Implemented AI-powered demand forecasting using Linear Regression with engineered time-series features (lag values, rolling averages, sales differences).- Designed a Genetic Algorithm-based slotting optimizer that allocates products to warehouse slots based on demand frequency, item compatibility, product dimensions, and proximity to dispatch zones.- Seamlessly integrated forecasting and slotting models into the Odoo ERP backend as custom modules with automated workflows.- Built and tested module functionalities using Katalon Studio to ensure reliability and accuracy.Results
- Achieved improved forecast accuracy compared to baseline methods, with Linear Regression outperforming alternative ML algorithms.- Optimized slotting significantly reduced average picker travel distance compared to other tested metaheuristic approaches across 3 different warehouse sizes.Tools & Technologies
- Python, Odoo, PostgreSQL, scikit-learn, pandas, FPGrowth, Genetic Algorithm, Linear Regression, PyCharm, Katalon Studio.
Churn Prediction for Telecom Customers
Problem
- Telecom companies face significant revenue losses when customers discontinue their services (customer churn).- Accurately predicting churn can help businesses identify at-risk customers and take preventive actions.Approach
- Data Preprocessing: Cleaned and transformed the dataset (encoding categorical variables, scaling, handling missing values).- Feature Engineering: Created meaningful features from customer demographics and service usage.- Modeling: Trained and compared multiple machine learning algorithms including Logistic Regression, Random Forest, and XGBoost.- Evaluation: Used accuracy, precision, recall, and F1-score to evaluate models, with a focus on recall.- Deployment: Built an interactive Gradio interface for real-time predictions and deployed it on Hugging Face Spaces.Results
- Logistic Regression achieved the best balance across metrics:- Accuracy: 80.6%
- Precision: 65.7%
- Recall: 55.9%
- F1-Score: 60.4%
- Random Forest and XGBoost performed slightly lower.Tools and Technologies- Python, Pandas, NumPy, Scikit-learn, XGBoost, Logistic Regression, Random Forest, Gradio, Hugging Face Spaces, Jupyter Notebook, VS Code.
Amazon Scraping Project
Problem
- E-commerce platforms like Amazon contain a vast amount of valuable product information, but accessing it manually is time-consuming and inefficient.- Businesses, researchers, and consumers often need structured data (e.g., prices, ratings, and product details) for market analysis.Approach
- Developed a Flask-based web application that enables users to search for products on Amazon and automatically scrape relevant details.- Leveraged Selenium and BeautifulSoup to handle dynamic page content, extract structured information, and provide real-time progress updates on the web interface.- Processed scraped data using Pandas and enabled CSV export for further analysis.Results
- Successfully scrapes product details such as title, price, rating, brand, category, sales rank, image, and product URL.- Provides a modern, responsive interface.- Enables users to download structured datasets (CSV) instantly.Tools and Technologies- Flask, Selenium, BeautifulSoup, Pandas, Webdriver Manager, PowerBI
Other Projects
Movie Recommendation System Using MovieLens Dataset
Problem
- Users needed better personalized movie suggestions.Approach
- Built a hybrid recommender combining collaborative filtering with content-based embeddings; deployed as an interactive Hugging Face app.Results
- Improved recommendation relevance for test users; quick, interactive demo showcases predictions.Tools
- Python, Pandas, scikit-learn, Hugging Face Spaces, Streamlit.
Brain Tumor Detection using Deep Learning Models
Problem
- Detect and segment tumors in MRI scans for research purposes.Approach
- Implemented and compared five deep learning models (AlexNet, MobileNet, VGG16, LSTM, Inception) to classify brain tumors from MRI scans.
- Preprocessed images, trained models, and visualized performance metrics to identify the most accurate classifier.Results
- Experimental results demonstrated that VGG16 model achieved the highest accuracy of 0.97, followed by AlexNet with an accuracy of 0.96.Tools
Python, TensorFlow, OpenCV, pandas, matplotlib, Google Colab.
2048 Game
Problem
- Implement classic 2048 game logic with a functional UI and efficient merging mechanics.Approach
- Programmed game logic, handled tile merging and score calculation, created user interface.Results
- Fully playable game with stable functionality.Tools
- C#, .NET (WinForms), Visual Studio.
Contact
Email:
[email protected]
Phone:
+971 50 871 7605