Healthcare

🍑 Project Context PeachyDay is a digital health startup helping patients manage chronic migraines. As a Data Scientist, I bridged the gap between raw data and business strategy by defining the product metrics framework and building the core ML features. 🛠 Technical Challenges Lack of Visibility: The team had raw data but lacked defined KPIs to measure product health or feature success. Clinical Validity: A “black box” model is useless in healthcare; predictions must align with medical understanding. User Retention: Users lacked motivation to log data daily without immediate feedback. 💡 My Solutions 1. Product Analytics & Metrics Framework I established the internal analytics system to track product health and user behavior. ...

🎯 Project Overview In Emergency Departments (ED), accurate prediction of resource utilization (like IV fluids) is critical for operational efficiency. Traditional models often ignore the rich information hidden in unstructured patient narratives (Chief Complaints). This project aimed to bridge this gap by developing a Multimodal Machine Learning pipeline that integrates structured clinical variables with NLP-derived text features. 🛠 Methodology Data Source Analyzed 13,115 patient records from the National Hospital Ambulatory Medical Care Survey (NHAMCS-ED). Input: Mixed data types including demographics (structured) and triage notes (unstructured). The “Early Fusion” Strategy I implemented an Early Fusion approach to combine distinct data modalities: ...