NewsPalette.com is a website I built to help users identify subjective language and propaganda in news articles. The goal was to promote media literacy by highlighting biased claims and emotionally charged phrasing across sources. As part of this project, I developed machine learning models to classify subjective statements and detect propaganda using both classic and modern NLP techniques. Some of my contributions include:

• building and training models using large language models (LLMs), conditional random fields (CRFs), and transformer-based architectures

• creating a custom web scraper that gathered and stored ~30,000 articles from 10 different news sites into a structured SQL database

• developing a full-stack web application that estimates the amount of subjective or propagandistic claims in articles, and allows users to browse and analyze articles with visualized predictions

The end product was an interactive tool for exploring bias in media content. My tech stack included:

• Python and it’s data science libraries (Pandas, Numpy, Scikit-learn)

• PyTorch

• Hugging Face Transformers

• PostgeSQL

• React, JavaScript, HTML, CSS

Unfortunately, I had to take it down shortly after joining Bloomberg due to potential conflicts of interest.