Samay

The measurement does not exist without the model. Sylvee, which the company describes as an AI wearable that monitors lung function at home (pulmonary volumes, trapped air, respiratory rate and breathing sounds) for early COPD detection, projects low frequency sound into the chest and reads the returning resonance, and a machine learning model turns that acoustic signal into lung structure and function measures and, in the company's 2025 published classifier, into a COPD diagnosis.

Unlike spirometry, which yields a directly interpretable measurement, the resonance signal means nothing clinically until the model interprets it, so the algorithm is the instrument and removing it leaves nothing to sell.

Clinician review is asserted and no mechanism is described. Results are surfaced through a companion app and a research portal for pulmonologists and primary care physicians to review, and the stated intent is to let clinicians intervene early rather than to automate a decision. Because the device is in development and not cleared, its output today sits in a research or monitoring context.

Nothing published describes alert thresholds for detected deterioration, who receives an abnormal reading and how quickly, or what happens when a reading is missed, which is the core question for a device positioned for continuous home monitoring.

The physical method is described in detail and the model in general terms. The company describes active acoustic resonance, a patented approach of projecting low frequency sound into the chest and reading the returning signal, combined with connected sensors, digital signal processing and machine learning.

Its 2025 abstract describes a classifier taking Sylvee data with age, sex and race, trained on 152 subjects and validated on 66, without naming the model type, the features extracted from the signal or the device version. No versioning or update practice is published, so a study sponsor cannot tell whether the model changed between visits.

Samay names no outside party in the architecture it describes. Data moves from the chest worn device to a companion app on the participant's own phone or tablet, then to a research portal and backend that share metrics with care teams, and the classifier described in the company's 2025 abstract was built by Samay's own staff, so no third party model provider appears in the published material. No hosting provider, subprocessor list or app infrastructure provider is named elsewhere either. Privacy terms on the site might name recipients. The acoustic data makes the chain matter, since it is not stated whether raw audio leaves the device or which parties receive it if so.

The evidence is published research short of independent validation. The company's 2025 abstract in the American Journal of Respiratory and Critical Care Medicine reports a COPD classifier on 218 subjects (107 COPD, 111 healthy, mean age 62), validated on 66 held out subjects: area under the curve 0.93, sensitivity 81 percent, specificity 83 percent against conventional pulmonary function tests.

An earlier study of 110 patients with a Florida pulmonary institute across COPD, asthma and healthy controls, also presented as an abstract, reported air trapping detection at 83 percent accuracy and COPD diagnosis at 90 percent against hospital pulmonary function tests, and the company has NIH small business support. Both are company led, conference abstracts rather than full papers, modest in size and on prototype or development devices, with no external validation cohort. The headline claim, early COPD detection that could replace effort dependent spirometry, needs larger, independent and prospective work to stand on.

The company's published abstracts and announcements carry no retention, training use or deidentification statement, though the privacy terms on its website may address them. The question specific to this device is audio. Sylvee projects sound into the chest and reads the returning resonance with speakers and microphones, is worn taped to the chest for continuous home use, and the company lists breathing sounds among its metrics.

A body worn microphone at home sits inches from the wearer's larynx and in the same room as the rest of the household. Whether that matters turns on where the signal is processed and what is kept: resonance features extracted on the device with the waveform discarded raise little, while raw audio streamed or stored for later analysis raises a great deal, including for people in the room who consented to nothing. No safety engineering is published.

The company publishes no HIPAA or business associate statement, and a privacy policy on the website is where one would most likely appear. The engagements described are research studies and endpoint capture for pharmaceutical trials, where the study protocol, institutional review board approval and the participant's informed consent govern and the sponsor or site holds the relationship with the participant; on that reading a vendor business associate agreement would not be the operative document, but that is an inference from the stage, and the company has not stated a scope position.

The boundary matters because the company's stated aim, continuous home monitoring for COPD, is care rather than research, and at that point the vendor would be handling identifiable health information on a provider's behalf.

There is no SOC 2, ISO/IEC 27001 or other attestation in the published material, unless the website carries a security page with one. The attack surface is broader than a software product's: a wireless body worn sensor paired with a consumer phone or tablet, a companion app on a device the company does not control, and a backend holding continuous physiological data for a clinically vulnerable population.

The company has appeared in at least one consortium listing under a different corporate name with the same founder and device, so an attestation could sit under either name. A pharmaceutical sponsor placing the device in a decentralized trial will require most of this contractually.

No FDA clearance is held: FDA's 510(k) database lists no clearance with Samay as applicant, and the company's own abstracts describe it as developing the wearable to facilitate COPD diagnosis. A device that measures lung function and diagnoses COPD sits within device regulation, so until it is cleared its use is confined to research, clinical trial endpoint capture and investigational monitoring, and every published accuracy figure is investigational.

The company's site describes Sylvee as an AI wearable that monitors lung function at home for early COPD detection; whether the investigational status appears beside that description is not clear. NIH small business funding is research support, not regulatory standing.

The company has published an aggregate evaluation and nothing by group, for a model that takes race as an input. Its 2025 abstract in the American Journal of Respiratory and Critical Care Medicine ("Diagnosing COPD With Sylvee", led by Samay staff with academic coauthors) describes a classifier built on Sylvee data and demographic information, specifically age, sex and race, trained on 70 percent of 218 subjects (107 with COPD, 111 healthy) and validated on the remaining 66, with no results by race, sex, age or body habitus.

Pulmonary function interpretation has moved away from race specific reference equations after finding they normalized worse measured lung function in some populations and delayed diagnosis, so a device positioned to replace spirometry that uses race as a model input needs its reason stated and its performance shown by group; neither is published. The split is described by subject counts, and the abstract does not say whether it was made by participant or by recording. No external validation cohort and no governance or model update policy is published.

The company has published an error rate for the product's diagnostic output with its method and denominator. Its 2025 abstract in the American Journal of Respiratory and Critical Care Medicine, whose corresponding author is at Samay, reports a machine learning classifier using Sylvee data with age, sex and race, trained on 152 of 218 subjects and validated on 66: area under the curve 0.93, sensitivity 81 percent and specificity 83 percent, with 25 of 31 COPD patients and 29 of 35 healthy individuals correctly classified (6 false negatives, 6 false positives) against conventional pulmonary function tests.

That is a result on a small internal validation set of a device the company describes in development terms, not performance in use, and it covers COPD versus healthy classification only, not the lung volume, trapped air or exacerbation monitoring outputs. No warranty, indemnity or remediation commitment is published, and nothing says how the device behaves on a poor signal.

What is described is a study architecture: the chest worn device, a paired mobile app and a research portal that moves signal from a participant to a research team. The company's abstracts and announcements describe no record system integration, standard or named deployment; an integration statement could still appear on its website.

For research use a portal is the relevant data path; for the continuous home monitoring the company intends, readings a clinician acts on belong in the record with provenance, timing and device identification, and the alerting question applies: who receives an abnormal reading, how fast, and what happens if nobody does.

The deployment surface is described and the hosting is not. Sylvee pairs wirelessly with the participant's own phone or tablet, a companion app handles signal analysis, and a research portal and backend carry respiratory metrics to care teams; the first two parts sit on hardware the company does not own.

Nothing published states where the backend runs, in which region, whether study data is kept separate per sponsor or site, or whether analysis happens on the phone, in the cloud or across both, which decides what leaves the participant's home. The website itself may say where the service is hosted. For decentralized trials, residency, sponsor separation and the chain of custody from sensor to case report form are regulated concerns.

No price is published, and the intended routes to market are discoverable: the company names three commercial contexts, early detection in primary care, continuous home monitoring, and high frequency endpoint capture in pharmaceutical clinical trials. Because the device is in development and not cleared, research studies and trial endpoint capture are the uses open to it now, and reimbursed clinical monitoring is not. No rate, study pricing or device cost is published.

Coverage is named and part of it is validated. The scope is respiratory, specifically COPD and small airway disease, across three intended settings: early detection in primary care, continuous home monitoring, and endpoint capture in pharmaceutical trials.

Validation stands behind the diagnostic use in a clinical setting: company led studies classified COPD against pulmonary function tests in 218 subjects (COPD and healthy controls) and in 110 patients at a Florida pulmonary institute (COPD, asthma and controls). Nothing published validates continuous home monitoring or trial endpoint use, and asthma and small airway disease have no separate results.