MatchA Social Media Website

Full-stack, Facebook-inspired social media site focused on real-time chatrooms and "matcha chats" between users.

Features:

    • Accessibility-friendly frontend design
    • SHA-encrypted user account system
    • User connection visualizer
    • Personal walls for users
    • Friend request functionality and notifications
    • MapReduce-style adsorption-based news article recommender
    • Socket-based, forum-style multi-user chatrooms

Technologies Used

JavaScript, Java, HTML, DynamoDB, Bootstrap, CSS, PUG, Apache Spark, AJAX, JQuery

Matcha

Matcha Chatroom

Matcha News Recommender and DMs

Matcha Sign-Up and Wall

Regional Restaurant Hunter

Location-based restaurant outing planner and cuisine-based restaurant recommender.

Features:

      • Google Maps API recommended restaurants visualization
      • SQL database-based preference filtering
      • Cookie-based "cart" of users' favorite restaurants
      • Graph-based comparator between user-selected restaurants
      • Overview of individual reviewers and their reviews
      • Random specification-limited restaurant recommender

Technologies Used

SQL, JavaScript, React, HTML, CSS

Restaurant Hunter

Restaurant Hunter Card

Random Restaurant

Restaurant Hunter Cart

Multilingual COVID-19 Twitter Sentiment Predictor Using Zero-Shot Modeling

Zero-shot modeling of a large Kaggle dataset of multilingual COVID-19-related tweets to predict user sentiment in service of future mental health detection.

Used a Recurrent Neural Network to predict sentiments like "sadness" and "anxiety" multilingually with around 70% accuracy.

Technologies Used

Python, Pandas, Matplotlib, SKLearn, Pytorch, Recurrent Neural Networks, Keras Long Short-Term Memory Architecture.

Sentiment Analysis

Anxious Example Tweet

Colinearity Heatmap