One picture from five signals.
The hard part is not capturing a single biomarker. It is reading several at once, in the noisy conditions of a home, and turning them into something a clinician can act on.
Fusing the streams.
Multi-stream input
A single smartphone session records voice, eye movement, facial expression, balance and neurocognitive performance, the raw material the model reads.
Multi-modal model
Rather than scoring each signal alone, the model reads all five streams together. The combination carries information no single stream holds on its own.
Improves with scale
The platform is designed to get sharper with every additional session. More data, validated against clinical outcomes, sharpens the risk picture over time.
A position that compounds.
Five validated streams, hospital-embedded validation and a growing longitudinal dataset are difficult to assemble and harder to replicate. As the dataset grows under clinical validation, the advantage widens.