Clinical Wearable Analytics for Digital Therapeutics
Large-scale CGM and activity analysis pipelines supporting digital therapeutics research.
Clinical wearable studies generate long, irregular, multi-source records per participant, and most of the analytical effort goes into making that data usable before any modeling begins.
With the AI Analytics group in the Digital Data Science department at Novo Nordisk, I focused on making that groundwork reusable rather than per-study.
Contributions
- Built reusable Python analysis pipelines for large-scale clinical wearable datasets covering 300+ patients with 30+ days of data each.
- Implemented feature extraction and aggregation workflows for continuous glucose monitoring and activity data to support downstream ML modeling.
- Developed statistical and ML-based analyses identifying key behavioral and physiological factors affecting glucose control, producing insights adopted by digital therapeutics teams.
Work conducted during an industry internship; described here at the level of publicly shareable methods and outcomes.