METRIA
- Completed
- 1,000+ automated tests
- 60+ recorded architecture decisions
Clinical system for nutrition medicine: the patient’s metabolic trajectory, the care plan and a governed clinical AI.
- Applied AI
- SaaS platforms
The problem
A nutrition physician sees 15 to 25 patients a day and has a few minutes to rebuild months of lab results, measurements and agreements before each visit.
Who it is for: Nutrition medicine practices and their patients.
What the system does
A canonical clinical timeline brings together visits, measurements, lab results and records. Care plans and care loops continue with the patient through the app. The AI-assisted summary cites the evidence behind every statement and requires medical review: nothing reaches the record without confirmation.
System screens
The screens stay in Portuguese, as in the system.
Modules
- Command Center
- Patient 360
- Clinical Story and Timeline
- Care Loop
- Clinical Plan
- Patient app (PWA)
- Clinical Intelligence Studio
- AI governance and operations
- Pix subscriptions
What the screen is made of
Every screen above depends on decisions like these, made in the code and covered by tests.
| Part | Decision |
|---|---|
| AI | A single gateway. Without an approved production release, the task answers as blocked. |
| Evidence | Every statement in the summary cites its source; anything without a source becomes a stated limitation. |
| Data | Internal boundary by default: an external provider only with a contract and no-training confirmation; outbound HTTP on a closed allow-list. |
| Isolation | 99 composite foreign keys including the practice make PostgreSQL itself enforce isolation. |
| Writes | Double clicks and retries receive the first response; a visit is edited under a row lock. |
Technology
- Laravel 12
- PHP 8.2
- PostgreSQL 16
- Blade
- Alpine.js
- Three.js
- Chart.js
- PWA
Integrations
- SyncPay (automatic Pix)
- Cloudflare Turnstile
- Private-network LLM (OpenAI-compatible API)
Stage
Completed
Completed. The screens show the verification environment, with a local test model and fictitious patients from the fixtures.
Want to see this system working?
We give a guided demo with demonstration data and talk about what it would solve in your operation.
Request a demo