Templates

Health Coaching

Reference architecture for an AI health-coaching assistant – multi-channel guidance grounded in clinical guidelines, with reminders and a strict safety posture, powered by the NAISH Hub.

A patient or a community health worker wants guidance over chat or voice, in their own language. This template shows a health-coaching assistant that answers from vetted guidelines, tracks the person over time, and sends reminders.

Health coaching reference architecture

How it works

  1. Channels – patients reach the coach over chat or voice; community health workers use an app.
  2. Gateway & consent – access, authentication, and explicit consent come first, before any clinical content.
  3. AI health agent (+ skills) – reasons about the request and calls skills: the model for language, retrieval over clinical guidelines for grounding, and the records service for continuity of care.
  4. NAISH Hub – serves the models on GPU. Sovereign AI core.
  5. Data & dependencies – a guideline vector store grounds answers, a patient record store holds history, identity verifies the person, and reminders go out through a queue to messaging.

Where the Hub adds value

  • Language & reasoning – Hub-served models coach in the patient's language.
  • Grounded, safe answers – retrieval ties responses to approved clinical guidelines; this is a high-stakes domain, so safety evaluation gates deployment. See Responsible AI.
  • Sovereignty – patient data and models stay on national infrastructure; see Governance & Risk.

Reusable building blocks

LayerThis templateSwap in for your context
ChannelsChat / voice + worker appAny patient or worker channel
Agent + skillsGuideline retrieval, records, remindersCondition-specific skills
Models (Hub)Local-language health LLMClinically fine-tuned models
ServicesRecords, notificationsWhatever platforms the programme runs
DataPatient records + guidelinesAny governed clinical source