Research Phase · Waitlist Open

Your heartbeat
knows your glucose.

Daibeats extracts glycemic intelligence from ECG signals using clinical-grade machine learning — validated across 131K+ ICU patients and wearable cohorts.

131K+
Patient Records
MIMIC-IV linked
48
Model Configs
SAGE-Net evaluation
3
ICU Databases
MIMIC-III/IV, eICU
5
Active Papers
IEEE · ACM · Elsevier

From electrical signal to metabolic insight.

A four-stage pipeline that converts 12-lead ECG data into clinically-stratified glucose estimates — no blood required.

01

ECG Capture

Standard 12-lead or single-lead ECG recording. No finger prick, no cannula — just electrodes on skin.

02

Feature Extraction

Multi-domain HRV analysis, morphological feature engineering, and temporal alignment against CGM reference windows.

03

ML Inference

Gradient boosted ensemble model trained on clinical-scale datasets. Validated against MIMIC-IV, AI-READI, and D1NAMO cohorts.

04

Glucose Estimate

Clinically-stratified glucose prediction with safety-aware confidence bounds. Designed to flag critical ranges.

Validated across

MIMIC-IVICU131K+
AI-READIT2D Cohort1,552
D1NAMOWearable9 subjects
PhysioCGMContinuousOngoing
eICUICU Multi-site200K+

From the lab notebook.

Be first to know when
Daibeats ships.

Whether you're a clinician, researcher, investor, or living with diabetes — there's a place for you on the waitlist.