Two markets. One shared flow.
The person records meals and progress. The professional publishes the plan. Only data authorised by the person moves between the two products.
Confidential · updated 29 July 2026
The product works.
The product works.
Now it needs distribution.
The app is available through an open test on Google Play.
The portal and backend are online.
Capital accelerates acquisition, sales and growth.
The opportunity
Value disappears
Value disappears
between appointments.
Consumer apps record daily life.
Practice software organises professional work.
NUTRIMI connects those two moments.
One data model, two products.
The consumer app and the clinical portal were not integrated after the fact: they share the same schema. The link is consent-based and granular, per data category, revocable from either side.
Language models now run on the phone.
An on-device model runs on the phone. It reduces cloud calls and variable costs. The free plan becomes more sustainable.
What is already built
Two usable products.
Two usable products.
One active infrastructure.
The app is available through open testing.
The portal verifies professional accounts.
The backend runs the active services.
Every figure comes from the repository and published systems.
387Dart filesMobile app code in lib/
1,080test cases899 Dart and 181 backend or rules
167unique active functions289 deployments across two regions
2,488offline foodsDatabase bundled with app and portal
91lessonsTen nutrition education units
The app
Plate recognition from a photo with component-level breakdown, nine meal-entry methods, a pantry with expiry dates and a shopping list shared across a household, a diary for water, weight, symptoms and activity, clinical pathways for sixteen conditions, pregnancy and lactation with a dedicated nutrition engine, a pre and post-operative bariatric pathway, ninety-one nutrition education lessons, an Experts tab pulling videos from twelve Italian nutrition channels, grocery reordering through Amazon and specialist shops, and full gamification with streaks, quests, leaderboards and an arena.
Offline-first on a local database
Optional cloud sync
GDPR export and deletion
The portal
Manual verification of professional registration, patient linking with per-category consent, a versioned meal-plan editor on a local database, pre-visit questionnaires pre-filled only from authorised data, private and immutable clinical notes, a calendar with patient confirmation, sixteen guided clinical checklists, AI draft generation with mandatory review, and token and cost tracking per professional.
Roles assigned only server-side
App Check enabled
No client access to health records
The part a fork does not copy.
Five hundred and forty-three original illustrations - mascot poses, lesson plates, notification art - drawn one by one, twelve product catalogues built per condition, a maternal nutrition reference matrix traced to its sources, reference guides for sixteen conditions with cited guidelines available in the app, and dietary restriction cards in fifteen languages with fixed, reviewed text. This is accumulated editorial and clinical work, not automated generation.
Defensibility
Four advantages
Four advantages
that widen over time.
Four choices strengthen the product over time.
Architecture, data, content and distribution work together.
On-device inference
A downloadable language model executed locally makes the free tier sustainable: a free user generates no inference cost. A competitor built entirely on the cloud pays for every free conversation and cannot lower prices without eroding margin.
Per-category consent
The patient-professional link is not one switch but four separate authorisations, revocable individually. That is a requirement you design at the start: retrofitting it means rewriting the permission model and the server surface.
Deterministic computation
Requirements, macronutrients and hydration are computed in code, not asked of a model. It is the only way to make someone accountable for the numbers, to keep them reproducible, and to avoid depending on an external vendor's behaviour.
Two sides, one channel
A consumer app on its own competes through paid acquisition. A practice tool on its own competes on price. Together they open a channel that needs no advertising budget: the professional brings their own patients, and patients look for a connected professional.
Monetisation
Revenue channels,
Revenue channels,
each with a declared status.
This table mirrors the internal revenue source of truth. A feature in the code is not revenue, and one stage does not imply the next. Last checked against the codebase, deployed Cloud Functions and production data on 24 July 2026.
| Revenue line | Status | Next step |
|---|---|---|
| Consumer subscriptions | Checkout live | Play catalogue connected and checkout reachable on all four plans: what it needs is volume at the top. |
| Free trial on the yearly plan | Ready | Publish the trial offer on the base plan in the stores: the code announces it automatically as soon as it exists. |
| Professional SaaS (AI Pro) | Ready for live | Stripe secrets to live mode and redeploy: the payment flow has already been exercised end to end in test. |
| Single consultation, €68 | Leads coming in | Requests come in and are recorded. Opening the one-off Stripe payment is a few hours of work. |
| Advertising (AdMob) | In production | Production ad units live and SSV deployed: revenue per user becomes measurable as the base grows. |
| Amazon affiliation | In production | Affiliate tag in the backend and links generated from the pantry: the first conversions get tracked. |
| AI call packs | Ready | App and server in place with idempotent crediting: the consumables remain to be published in the stores. |
| Coins and in-app shop | Lever available | Coins are earned only today. Making them purchasable is a lever that is ready, held back as a product choice. |
| Marketplace, corporate welfare, licensing | To validate | One-shot programmes, marketplace, corporate welfare and API licensing: natural expansions once the base is consolidated. |
Structurally low acquisition cost
The free plan for professionals is not a timed trial: it is a channel. Every professional who links their patients brings users into the app with no advertising budget, and every patient looking for a connected professional pulls from the other side. That is why the two products were built together.
Margin protected by on-device inference
A conversational assistant built purely on the cloud pays for every free conversation. Here the model runs on the device: the marginal cost of a free user tends towards zero, and pricing can stay aggressive without eroding margin as the base grows.
Funnel instrumented end to end
The subscription screen appears at three relevant moments, capped at three prompts per user with seventy-two hours between them; the free plan carries a monthly quota of AI meal plans; return reminders fire at three and seven days of inactivity. The funnel records views, checkout, exit, trial start, conversion and subscription end: from the first user onwards, conversion, retention and margin can be read by channel, not estimated.
Revenue model
Two markets, one product to maintain.
Development cost is shared across both sides: the same food database, the same plan model, the same backend. Every clinical improvement serves both audiences.
Mobile subscription
Premium at €9.99 a month or €59.99 a year. Premium with a consultation at €84.99 a month or €144.99 a year. The free plan stays usable and includes the local model with no limits, which keeps the cost of a non-paying user low.
SaaS on Stripe
Free practice plan up to five patients; AI Pro at €34.90 a month or €349 a year up to one hundred patients with sixty monthly AI drafts. The free tier is not a timed trial: it is an acquisition channel that brings patients into the app.
Advertising and affiliation
Advertising on the free plan, rewarded video to extend cloud calls, affiliation on the pantry and shopping list with explicit disclosure. None of these lines has produced revenue yet.
What we have not written
One-shot programmes, a marketplace, corporate welfare and API licensing exist only as hypotheses in a strategy document. There is no code, and there will not be until there is commercial validation.
Roadmap
Where the capital goes.
Not into development: the product exists. Into switching on channels that are already built, and into acquisition from the professional side, which is the cheapest one.
Done
A complete product on two sides
The app is available through an open test on Google Play. The portal verifies professional accounts. The serverless backend is active. Legal and privacy documents are versioned and accepted server-side.
Done · July 2026
An observable funnel
Automatic subscription prompts at moments of value, a free quota on AI meal plans, return reminders and full recording of the subscription lifecycle.
In progress
Switching on payments
Stripe in live mode for the professional SaaS, trial offer and consumables published in the stores. Configuration steps on code that is already deployed and exercised: days, not months.
Next
First measured cohort
With the funnel already instrumented, conversion, retention and margin can be read by channel from the very first users. Marketing investment decisions start from numbers rather than assumptions.
Then
Scale from the professional side
Every active professional brings their own patients along. Once that coefficient is measured, acquisition becomes an investment with a computable return instead of an advertising bet.
Let's talk
The product can be tried today,
The product can be tried today,
from both sides.
Download the app, open the portal, connect a patient: the entire flow can be completed without paying anything and without a guided demo. Detailed materials are available on request.