Use cases

Where the inference layer fits.

Patterns for the teams building on Medera: what each integration sends, what comes back, and the limits to respect. Your product owns capture, consent and the decision; Medera measures and infers.

01

Measurement-based care between visits

See who has changed since the last visit, against their own history, before the next appointment.

/v1/analyze · /v1/inference

You send

  • PHQ-9 and GAD-7 at each contact
  • Sleep and heart-rate variability from a wearable
  • The patient's earlier sessions as the baseline

You get back

  • A stable, watch or review tier per patient
  • Which RDoC domains moved, and by how much
  • Post-test probabilities with their priors
  • A safety block that reaches a clinician

Respect the limits

Tiers are a reason to look, not a decision. Keep the clinician's review in the loop and document it.

02

Digital mental health programs

Add objective, explainable measures to an app's check-ins without ever sending a face or storing a voice.

/v1/facial/analyze · /v1/vocal/analyze · /v1/inference

You send

  • Blendshape scores from the front camera, computed on the device
  • A short voice check-in as a WAV file
  • In-app questionnaires

You get back

  • Expression, gaze and blink measures
  • Pitch, voice quality and timing
  • Scored instruments and language markers
  • Change against the user's own baseline

Respect the limits

Do not present a biomarker, tier or probability to a user as a diagnosis, and route safety blocks to a person.

03

Clinical research and trials

Exploratory digital measures with documented, versioned methods that a protocol can cite and an auditor can reproduce.

/v1/analyze · /v1/biomarkers

You send

  • Study recordings and wearable exports, per visit
  • Per-participant baselines
  • Instrument responses at protocol time points

You get back

  • Deterministic values with an engine version on every response
  • Quality flags that make missingness explicit
  • Within-person change with stated error rates

Respect the limits

These are research measures. Validating one as an endpoint for a given population is the study's work, not a property of the API.

04

Telehealth and conversational care

Measure how a conversation goes, not what is said: timing, voice and response latency from two-channel session audio.

/v1/vocal/analyze · /v1/clinical-context/analyze

You send

  • Session audio with one speaker per channel
  • An English or Arabic transcript, when you have one
  • Channel roles, for example clinician and patient

You get back

  • The vocal panel for each speaker
  • Response latency between speakers
  • Speaking rate and language markers from the transcript

Respect the limits

Recording a session needs the consent your jurisdiction requires. The API keeps nothing, but your capture layer does.

Tell us what you are building.

Sandbox today. Production under agreement.

Or make a first request with the sandbox key.