Technology

Sensing approach

The Awareble wearable captures autonomic, metabolic and inflammatory data continuously and passively on the body. The device is designed for years of uninterrupted operation.

The approach is reagent-free. Historically, the primary barrier to long-term continuous sensing has been that chemical reagents degrade, drift and require frequent replacement. Our sensing architecture avoids this entirely. The signal processing that makes this possible was not tractable with classical analytical methods until recently. Current computational approaches have changed that.

Individual physiological modelling

The sensing is one half of the proposition. The other half is what we do with the continuous data.

Rather than comparing measurements to population reference ranges, we build a compressed representation of each individual's physiological state that evolves with incoming data. We learn how the body reacts and interacts across domains, and what constitutes a meaningful deviation for that specific person.

The computational approach originates from work in medical imaging, where we developed methods for compressing variable-resolution temporal sequences into clinically useful representations. The same principle applies here: take a long, multi-dimensional physiological signal and compress it into something that retains clinical relevance while discarding noise.

Over time, the profile captures resting baselines, circadian patterns, exertion responses, recovery dynamics. Deviations from an individual's own trajectory carry more clinical weight than deviations from a population distribution, because they reflect actual change rather than statistical distance from a group mean.

Early validation

In our first prototype validation study (April 2026), we tested whether the sensing approach can reliably distinguish between chemically complex fluids. Across 114 blinded measurements spanning 13 distinct fluids, the system achieved 91.2% classification accuracy (5-fold cross-validated), with nine fluids identified at 100% sensitivity.

An unprompted finding was that the model, trained only to identify fluids, also encoded concentration information in its learned representations. For four fluids, R² exceeded 0.80 on held-out data, meaning the same sensor architecture can simultaneously answer what is present and how much. These results confirm the feasibility of the core sensing principle.

Why multi-domain matters

Single-domain wearables exist. Heart rate monitors capture autonomic data. Continuous glucose monitors track one metabolite. What none of them do is capture multiple physiological domains simultaneously and interpret them as a connected system. The relationships between autonomic, metabolic and inflammatory signals contain information that no single domain provides on its own. Combining them continuously, at the individual level, is what opens up the clinical applications described on our main page.