that-tech-geek/digital-twin

This research prototype turns longitudinal glucose and related signals into patient-specific glucose forecasts, hypoglycemia-risk estimates, and explicitly hypothetical counterfactuals. The principal workflow is source adaptation, canonicalization and timeline preparation, causal features and future-event labels, patient-level validation, model training, and three-layer twin simulation; evaluation and artifacts capture research results. The dashboard also exposes state, uncertainty, risk, and scenario simulation. Links follow the README’s materialized architecture and sampled calls; unsampled subsystem wiring is shown only where documented.