The project describes sensing inputs, model cadence, clinical constraints, and how alerts should be calibrated for trust.
The design treats this work as a signal-monitoring problem with streaming traces and human-facing intervention points.
The methods narrative emphasizes the practical gap between a predictive model and a system a patient could actually live with.