RANEEM MOUSA
MATH AND COMPUTER ENGINEERING STUDENT
Projects
A collection of my recent work in AI, software development, and engineering.
Virtual Fitting Room
An AI-powered virtual fitting room that allows users to try on clothes virtually. Built with computer vision and machine learning to provide a realistic and interactive shopping experience. Features include real-time body measurements, clothing visualization, and personalized recommendations.
Route N Roam
A Java-powered, location-based recommendation engine that blends Tinder-style swiping with Yelp review data and a Pinterest-style feed—letting users discover local restaurants, activities, and self-care spots, save favorites to a wishlist, schedule visits, read reviews, and find and connect with others planning to go to the same place at the same time—all within one intuitive interface.
ProfBotix: Math and Programming Tutor Chatbot
Flask web app for math and CS tutoring with step-by-step solutions, practice quizzes, OCR-based problem parsing, and OpenAI-powered contextual explanations.
Online Dresses Shopping Cart & Wishlist
A C++ console-based application simulating an online fashion boutique. Users can add dresses, skirts, tops, and accessories to a shopping cart or wishlist, adjust quantities, and view real-time total cost calculations. The implementation showcases robust object-oriented design—utilizing inheritance and polymorphism—and includes a custom linked-list for dynamic item management.
Spotify Top Songs Analysis
A Streamlit-powered web app that taps into the Spotify API (via Spotipy) to pull in your most-played tracks and surface their audio features—danceability, energy, valence, and more—through interactive charts. Analyze your listening habits at a glance and get AI-driven song and artist recommendations tailored to your personal taste.
NdLinear vs nn.Linear Performance Comparison
Implemented a "NdLinear" layer to preserve multi-D input structure (text & video) and compared it with a conventional flatten+`nn.Linear` on AG News (text) and UCF-101 (video). Across both tasks—even when integrated with a frozen ResNet-18 backbone—NdLinear achieved lower validation loss and higher accuracy without adding parameters or slowing down training.
Contact
Let's connect and discuss how we can work together.