Data Science Analyst, Master's Intern
Suade
Apr 2025 - Jun 2025 (3 months)
Built a multimodal TikTok analysis pipeline combining audio, speech-to-text, video emotion, on-screen text, and LLM sentiment, raising comment-level sentiment accuracy from -50% to -80% and replacing single-signal analysis.Preprocessed and standardized TikTok dataset cutting modeling time by -33% (6 hours to 4 hours), reducing compute cost and noise, and enabling more consistent KPI rollups and dashboard insights.Launched a Streamlit dashboard turning multimodal signals into actionable insights, ranking influencers, identifying content gaps, optimizing posting windows, and providing a single sentiment KPI per video to guide marketing strategy.Key Skills and Deliverables: Analytical Skills, Problem-solving Skills, Tableau, Statistical Method