VisualGPT: Multimodal AI for User Perception
Developed dual-route model integrating language models to predict user impressions on visual content, improving accuracy by +20% over baseline models.
Welcome to my data science and computational social science projects. I specialize in analyzing large-scale datasets, building predictive models, and applying machine learning to understand human behavior and social phenomena.
Developed dual-route model integrating language models to predict user impressions on visual content, improving accuracy by +20% over baseline models.
Analyzed 20M+ multilingual social media posts using advanced NLP pipelines to track misinformation spread and predict viral content with 85% accuracy.
Trained language models on 3.5B tokens using HPC to analyze 70+ years of news articles and track moral attitudes across cultures.
Designed deep neural networks for longitudinal health data analysis, achieving 4.09 years mean absolute error on aging prediction with 2,000+ participants.
Developed comprehensive educational resources and hands-on tutorials to teach natural language processing techniques to psychology researchers.