Live · v2.4 · HDS-conform · clinician-in-the-loop

A clinical vertical LLM // 2026

The engine
that turns any
content into care.

Analyzing 48s of patient voice · FHIR context loaded

Signal · input

🎙 "I feel short of breath after climbing stairs."

Literacy · matched

Grade 6 · ES-MX

Safety layer

3 rules passed · clinician approved

Delivered

Micro-lesson · 90s audio · quiz

Book a 20-min demo → See the platform ⌘K Talk to Saana

−30%

Avoidable readmissions, at pilot sites with chronic-care cohorts.

15+ languages

Cultural & literacy adaptation — from Grade 3 to clinical register, per patient.

<90s

From PDF or discharge note to personalized lesson, audio summary & quiz.

100%

Clinician-verified. HDS-conform. Full audit trail, every output sourced.

02 · For whom

One engine. Two sides
of the same care loop.

Saana plugs into the clinician's workflow and the patient's daily life at once — providers keep control and context, patients get education that actually adapts.

Providers · clinicians · payersA

Give your team an engine,
not another dashboard.

Upload guidelines, discharge notes or existing materials. Saana turns them into compliant, multi-modal education — auto-routed to the right patient at the right moment.

✓FHIR-native · drops into Epic, Doctolib, Maiia, HDS-hosted EHRs.
✓Full audit trail · every output is sourced, versioned, clinician-signed.
✓Real-time gap detection — know who's falling behind, and why.

Cohort · Post-MI recovery124 patients

Engagement (7d)87%

Flagged for clinician review6

Patients · members · caregiversB

Care that listens,
in plain language.

A voice-first assistant that meets people where they are — their language, their literacy, their context. No more generic PDFs. No more guessing with ChatGPT.

✓Voice, text, audio, or written — whatever works today.
✓Adapts to reading level, culture & EHR context in real time.
✓Escalates to a human clinician the moment it should.

Hi Maria — your cardiologist shared your discharge plan. I can read it out loud, or answer any question. Would you like a 2-minute audio summary?

Sí, por favor. Y explícame lo de la aspirina.

Claro. 🔊 Reproduciendo… "La aspirina a dosis baja ayuda a…"

03 · How it works

Any content in. Personalized
care out. In under 90 seconds.

Step 01 · Ingest

Drop any asset you already have.

PDFs of protocols. Discharge notes. Explainer videos. Audio from a clinician. Saana parses them, normalizes them, and links them to the stateful patient graph.

PDFACC/AHA Hypertension guideline.pdfParsed · 47 concepts

MP4Post-op physio intro.mp4Transcribed · FR

M4ADr. Alami · 02:14Processing…

IMGECG-2026-04-18.pngLinked · patient #482

Step 02 · Engine

The Saana engine — a stateful clinical stack.

Not a generic chatbot. A vertical architecture built for clinical ambiguity, with deterministic guardrails.

Stateful patient graphFHIR · Persistent

Agentic reasoningContext RAG · Clinical logic

Safety layerRule-gated · Zero-trust

Step 03 · Deliver

One upload → the right format for the right person.

Every output is sourced, clinician-approved, and adapted to the patient's literacy, language and cultural context — across any channel.

Audio summary · 90sES-MX · Grade 6

Interactive quiz · 4 qFR · Grade 9

Micro-lesson · 2 minAR · Grade 5

Smart reminder · SMSEN · 08:00

Clinician brief · PDFEN · Clinical

04 · Under the hood

A vertical LLM for clinical work — not a wrapper on a general model.

01 / Stateful Patient Graph

Memory that compounds, per patient.

A persistent, FHIR-compliant graph capturing clinical state across every touchpoint — so every answer is contextual, not generic.

FHIR-native Persistent Compounding

02 / Agentic Reasoning

Specialized agents for clinical ambiguity.

A system of specialized agents — triage, education, adherence, escalation — that handle the messiness of real medical workflows.

Context-aware RAG Clinical logic Dynamic routing

03 / Safety Layer

Deterministic guardrails. Clinician-in-the-loop.

A rule-based gate sits between model output and patient. Every delivery is sourced, traceable, and signed — hallucinations never reach the patient.

Rule-gated Zero-trust HDS-conform

The learning loop

Every interaction makes the next one better — for every patient.

Virtual patient personae, generated from the stateful graph, are used in reinforcement testing — so real patients never bear the cost of system improvement.

05 · Outcomes that compound

We don't sell content.
We sell behavior change.

Per avoided readmission

$12–17^K

Saved per case, per hospital — compounding across insurer contracts and patient cohorts. Pilot evidence across chronic-care programs.

Source · PMC/NCBI · 2024

−30%

Reduction in avoidable readmissions across pilot chronic-care cohorts.

3.4×

Patient engagement vs. static PDF / video education baselines.

15+

Languages and literacy levels — delivered from the same upload.

07 · FAQ

What clinicians
ask us first.

Q · 01

Does Saana replace my EHR or workflow tools?

No. Saana sits alongside your EHR. We plug into FHIR-native systems (Epic, Doctolib, Maiia and others) and route outputs through your existing channels — SMS, patient portal, app. Your workflow stays intact; patient education becomes continuous.

Q · 02

How do you prevent hallucinations from reaching patients?

Every output goes through a deterministic, rule-gated safety layer, is traced to a sourced passage, and can be clinician-signed before delivery. The model never speaks directly to the patient without guardrails.

Q · 03

Is my data used to train general models?

Never. Saana is HDS-conform, hosted on certified infrastructure in your region. Patient data stays in the tenant. No training on customer data without explicit, contract-level consent.

Q · 04

How long does onboarding a clinic really take?

A typical pilot goes live in 2–4 weeks: 1 week of content ingestion and persona mapping, 1 week of clinician review, and a staged rollout to patients. No custom integration is required to start.

Q · 05

What outcomes can we realistically measure in 90 days?

Engagement rates, knowledge retention (quiz-derived), adherence proxies, and time-to-first-follow-up. On chronic cohorts, we've observed measurable shifts in avoidable readmission risk within a single quarter.

The engine
that turns any
content into care.

One engine. Two sides
of the same care loop.

Give your team an engine,
not another dashboard.

Care that listens,
in plain language.

Any content in. Personalized
care out. In under 90 seconds.

Drop any asset you already have.

The Saana engine — a stateful clinical stack.

One upload → the right format for the right person.

A vertical LLM for clinical work — not a wrapper on a general model.

Memory that compounds, per patient.

Specialized agents for clinical ambiguity.

Deterministic guardrails. Clinician-in-the-loop.

Every interaction makes the next one better — for every patient.

We don't sell content.
We sell behavior change.

What clinicians
ask us first.

Turn your next PDF into the
best care conversation your patients
have had all year.

Tweaks ⌥T

The engine that turns any content into care.

One engine. Two sidesof the same care loop.

Give your team an engine,not another dashboard.

Care that listens,in plain language.

Any content in. Personalizedcare out. In under 90 seconds.

Drop any asset you already have.

The Saana engine — a stateful clinical stack.

One upload → the right format for the right person.

A vertical LLM for clinical work — not a wrapper on a general model.

Memory that compounds, per patient.

Specialized agents for clinical ambiguity.

Deterministic guardrails. Clinician-in-the-loop.

Every interaction makes the next one better — for every patient.

We don't sell content.We sell behavior change.

What cliniciansask us first.

Turn your next PDF into thebest care conversation your patientshave had all year.

Tweaks ⌥T

The engine
that turns any
content into care.

One engine. Two sides
of the same care loop.

Give your team an engine,
not another dashboard.

Care that listens,
in plain language.

Any content in. Personalized
care out. In under 90 seconds.

We don't sell content.
We sell behavior change.

What clinicians
ask us first.

Turn your next PDF into the
best care conversation your patients
have had all year.