CUSTOM CONTENT & AI FINE-TUNING

Start with 964 evidence-based lessons. Add your voice on top.

Digital Wellness Academy ships with a canonical library built from deep research and clinical synthesis. You don't start empty. You build on a foundation that already works.

And every practice has more to add. Protocols refined in supervision. Lectures recorded over a decade. Patient handouts tested across hundreds of cases. The ingestion pipeline lets you layer your clinical voice, your practice protocols, and your teaching materials onto the canonical base - and fine-tune a practice-specific adaptive model on the combined corpus.

Your practice's content, on top of our foundation

The canonical 964 lessons are built from deep research, clinical synthesis, and the Five Pillars framework - so every licensee ships with a full working library from day one. On top of that, every practice has clinical material the platform can fold in: protocols refined in supervision, lectures recorded over a decade, patient handouts tested across hundreds of cases. Material that would otherwise sit in folders, hard drives, and session memory.

You've written dozens of handouts

Patient education PDFs, intake sheets, skill-building worksheets - often authored over a decade of practice. They live in a folder labeled handouts and get emailed one at a time.

You've recorded lectures and workshops

Grand rounds, CME talks, community workshops, YouTube uploads. Hours of recorded expertise. Zero of it is indexed, transcribed, or assignable to a patient between sessions.

You have clinical protocols

The specific way your practice approaches panic disorder, OCD exposure, trauma processing. Documented in Word, refined in supervision, never operationalized as patient-facing learning.

You have a teaching voice

The way you explain catastrophic thinking, exposure hierarchies, behavioral activation. It's why patients refer their friends. It dies in the session room.

The platform's canonical library is the floor. Your material is the layer that makes the library yours. The ingestion pipeline folds your content in alongside our 964 lessons - measured against the same outcome framework, shipped on your domain, in your voice - so the combined library reflects both the canonical evidence base and your clinical specificity. Neither alone is as strong.

Multimodal ingestion

The platform accepts content in whatever format you already have it. We do the conversion work - so you don't have to reformat anything to get started.

Documents

PDF, DOCX, Markdown, HTML, EPUB. OCR for scanned pages. Preserves structure (headings, lists, tables, images).

Clinical protocols, handouts, research papers, training manuals, intake forms.

Video

MP4, MOV, YouTube / Vimeo URLs. Auto-transcription generates searchable transcripts, captions, and time-coded chapter markers.

Recorded lectures, CME talks, patient education videos, demo walkthroughs.

Audio

MP3, WAV, podcast RSS. Transcribed with speaker diarization where applicable. Listenable alongside a rendered text version.

Podcasts, dictated notes, meditation recordings, clinical interviews.

Images

PNG, JPG, SVG, WebP. Clinical diagrams, infographics, anatomical illustrations. Indexed with descriptive alt-text for accessibility + SEO.

Anatomy diagrams, decision trees, infographic handouts, practice branding assets.

Structured data

CSV, JSON, assessment schemas. Imported as platform-native assessments or tracking logs. Outcome data pre-populates trend charts.

Custom assessments, outcome datasets, tracking checklists, intake scoring.

Web content

URL crawling of your existing blog, clinic website, patient-education pages. Extracted, normalized, and made assignable inside the platform.

Existing blog posts, practice website pages, Notion / Confluence wikis.

The ingestion pipeline, step by step

From upload to patient-facing lesson, with a clinical-review checkpoint you control. No black boxes.

1
Upload
Dashboard, API, drive integration, URL
2
Extract
OCR, transcription, parsing, normalization
3
Classify
Auto-suggest track, pillar, condition tags
4
Review
Your clinical lead approves or edits
5
Publish
Goes live in your practice library
6
Fine-tune
Opt-in to queue for model training
Diagram placeholder - to be replaced with a polished NotebookLM visualization.
1. Upload

Drag-and-drop through a web dashboard. Or hit a REST endpoint from your EHR. Or authorize a Google Drive / Dropbox / Box folder and we watch it. Or paste a URL and we crawl a page. Bulk upload of entire handout folders is a common first move.

2. Extract & normalize

PDFs are OCR'd when needed. Videos and audio are transcribed with clinical-vocabulary fine-tuning. Documents are parsed with structure preservation (headings, lists, tables). Media is re-encoded for streaming. Nothing is lost in the conversion.

3. Classify & tag

The platform auto-suggests a track (e.g. Anxiety & Fear), a course (e.g. Panic Disorder), and pillar alignment (e.g. Cognitive Flexibility). You accept, adjust, or split a single source into multiple lessons.

4. Clinical review (your step, not ours)

Your clinical lead sees the ingested, tagged draft and approves, edits, or sends back. The platform never publishes provider content without a named reviewer signature. This is the gate that keeps quality at the standard of the canonical library.

5. Publish to your practice library

Your content appears alongside canonical lessons in your patients' learning path. Assignable from the provider portal. Trackable against outcomes. Indexable in your practice's public SEO surface if you mark it public.

6. Fine-tune queue (optional)

If you opt in, your content joins the fine-tune corpus for your practice's adaptive models: recommendation engine, AI coach voice, content atomizer. All fine-tuning is tenant-scoped unless you explicitly opt into cross-practice aggregate sharing.

THE INFLUENCE MAP

Every signal that shapes the model - including yours

The adaptive model isn't trained on one source. It fuses clinical expertise, learner signals, community activity, and scientific literature into a practice-specific intelligence that reflects how your practice teaches.

Inputs
Your voice
Clinical expertise
  • • Practice protocols
  • • Provider profiles
  • • Lecture + video library
  • • Patient handouts + SOPs
  • • Supervision frameworks
Learner signals
  • • Quiz performance
  • • Lesson drop-off
  • • Engagement velocity
  • • Star ratings + verbatim
  • • Time-on-task patterns
Community signals
  • • Forum questions + themes
  • • Peer-to-peer activity
  • • Cohort-level patterns
  • • Public-lesson social shares
  • • Common misconceptions
Scientific foundation
  • • Canonical 964 lessons
  • • 22 validated assessments
  • • Outcome trajectories
  • • Ongoing literature review
  • • Peer-reviewed citations
Your practice's adaptive model
Retrieval · Recommendation · Generation

A fine-tuned instance, scoped to your tenant, re-trained on your latest signals. Every week, smarter about how your patients learn and what your practice teaches.

Outputs
Personalized path
Course sequencing, pacing, and lesson recommendations per patient
AI coach voice
In-lesson support tuned to your clinical register and vocabulary
Atomized content
Blog + email + social drafts in your practice's tone
Clinical dashboards
Trends, outcomes, and at-risk signals for your provider team
Diagram placeholder - can be replaced with a polished Claude / NotebookLM visualization when ready.
THE PROVIDER-VOICE SPECTRUM

You can be anywhere on this spectrum - and move along it as you grow.

Some licensees adopt the canonical content as-is and let patients benefit from the unified library. Others inject their full clinical voice - practice protocols, provider profiles, lecture series, supervision frameworks - and watch their patients learn from their practice's mind. Most land in between. Provider voice is not an extra-cost add-on. It is the point.

Low effort
Adopt canonical content as-is. Patients benefit from the unified 964-lesson library and the shared evidence base.
Moderate
Ingest a few signature protocols and recorded lectures. Auto-tune the AI coach tone to your clinical writing.
Deep
Full provider-profile injection: lecture catalog, supervision framework, practice-specific onboarding. Every output reflects your practice.

What gets fine-tuned - and what doesn't

  • Tuned: your content retrieval weights, your recommendation priors, your AI coach language, your atomizer voice.
  • Not tuned by default: the MAIA distress classifier. Safety-critical models stay on the canonical-data training set unless you explicitly opt into practice-specific adaptation with de-identified aggregate review.
  • Never tuned: models shared across tenants. Your corpus trains your adaptations, not anyone else's.

Privacy, tenancy, and IP governance

Your content is your content. The ingestion pipeline is designed around that principle - not bolted on.

Tenant isolation

  • Your content lives in a tenant scoped to your practice. Row-level access in the database, not application-layer filter.
  • Other practices cannot see, search, or recommend against your content under any circumstance.
  • Cross-tenant analytics are de-identified aggregate rollups. A practice does not show up in another practice's usage data.

IP ownership

  • You retain all IP rights to content you upload. The platform's license is limited to hosting and serving it to your patients.
  • If you end your license, you get a structured export of everything you uploaded plus the derivative lessons, assessments, and fine-tune checkpoints.
  • Any cross-practice value capture (e.g. marketplace licensing of your content to other practices) is opt-in, priced, and revenue-shared.

HIPAA posture

  • Same BAA and infrastructure posture as the rest of the platform: tiered-access privacy, audit logging on PHI-adjacent writes, BAA-compliant cloud deployment as a prerequisite for production licensing.
  • Patient-identifiable content uploaded as part of ingestion (e.g. case reports) is flagged and reviewed for de-identification before publishing.

Clinical governance

  • Every ingested item passes through your practice's clinical-review step before going live. Named reviewer, audit log, version history.
  • Retracting a lesson unpublishes it and archives it - patients who had started it get an appropriate transition, not a broken link.

Integrations with the tools you already use

Most practices don't want another upload dashboard. Connect what you already have.

Cloud drives
Google Drive, Dropbox, Box, OneDrive
Video platforms
YouTube, Vimeo, Wistia
EHR (roadmap)
FHIR endpoints, Epic App Orchard, Cerner
Knowledge bases
Notion, Confluence, practice wikis
Podcasts
RSS feeds, Apple Podcasts URLs
Direct API
REST + webhook for custom pipelines
Web crawl
Your blog, website, practice pages
Manual
Drag-and-drop web dashboard

Get a content-ingestion consultation with your beta application

Beta licensees get a 30-minute session with our clinical content team to inventory what you already have, map it to the platform's tracks and pillars, and scope the ingestion work. No content required before the call - bring a list of what exists, we handle the migration.

More from our network

🚀SoloFrameHub

Two AI-native academies for founders. GTM OS is the go-to-market operating system you run your day in; 60-Day Founder turns an idea into investor-ready proof.

🧠SVTech AI

Build vertical AI products with platform discipline.

🩺SVTech Digital

The operating system for psychiatric practice growth.

💼BizFrameHub

BizFrameHub: practical AI for professionals. TradesOS: run a more profitable trade business.

🗣️SVTech English

Master business English for tech professionals.

🇨🇴Colombia Conversa

Master Colombian Spanish. Discover Bogotá.