Photoplethysmography (PPG) from the camera
We extract pulse waveform features from fingertip video. Signal quality is scored on every reading so low-confidence results are visible, not hidden.
Science and validation
How we build models, what we publish, and how we invite participation — without inventing credentials we do not have.
Methods
Self-supervised representation learning
Models learn from large amounts of unlabelled physiological signal before fine-tuning on smaller labelled sets — reducing overfitting to a single lab cohort.
LR-HSMM for temporal structure
Long-range hidden semi-Markov models help separate stable baseline shifts from short-lived noise when building personal trends over weeks.
Public literature we build on
Our validation approach
Field performance matters more than a single lab number. When our validation cohort is complete, we will publish stratified results on the accuracy page: error bands, coverage (how often a recording yields a result), and breakdowns by skin tone, motion, and heart-rate range — with methodology and limitations spelled out. How each vital is measured is on Technology.
Until those tables exist, we show signal quality inside the product instead of quoting headline percentages that hide uncertainty.
Scientific advisory board
We do not list named advisors until real people have agreed to be named. A wellness product that invents scientists on its website is worse than no section at all. When our advisory board is in place, names and affiliations will appear here — not logos, not stock photos.
Research program
Optional research consent in onboarding lets you contribute de-identified metrics for validation studies. Raw video never leaves your phone; research uploads are separate and revocable.
Publications and preprints will be linked here when available. There is nothing to cite yet — we would rather say so than pad the page.
© 2026 EDMC. All rights reserved.