Contributions to LifeBell AI
- ● Directed a cross-functional team of data scientists and nurses, overseeing the collection, storage, and labeling of a large ground truth dataset for algorithm validation in LifeBell’s v1 FDA submission
- ● Spearheaded the creation of data science and data engineering infrastructure, emphasizing data engineering best practices for working with columnar data, implementing APIs that streamlined the collection, mapping, and labeling of high-frequency sensor and EHR data, reducing processing time and improving team productivity
- ● Designed and deployed a robust signal processing-based respiration rate algorithm for LifeBell’s v1 FDA submission, achieving a Mean Absolute Error (MAE) below the FDA’s precedent of 3 breaths per minute
- ● Developed and deployed a posture and activity recognition convolutional neural network (CNN), enabling real-time characterization of patient behavior with a validation f1-score of 97%
- ● Delivered actionable analytics to product and operations teams, providing critical insights that informed the decision-making process for ongoing clinical trials