FitZen
AI-powered diet and nutrition tracking.


- Project
- Our own product
- Our role
- Product · Mobile development · Backend · Release
- Duration
- 5 months
01 / PRODUCT & APPROACH
From a meal to an everyday habit.
FitZen is our own product for making nutrition tracking part of everyday life. Photo-based meal analysis, daily tracking and recipe creation come together in a connected experience on iOS and Android.
Nutrition tracking turns into a chore and gets abandoned when meal logging and planning are scattered across screens. AI can ease that, but with two traps: presenting model output as a health claim, and depending on the model itself. When a provider retires a model version, every released build of the app breaks at once.
The daily experience
- 01
Capture a meal
Start with a food photo from the camera or gallery.
- 02
Review the result
Review estimated foods, portions and nutrients, then add them to the daily log.
- 03
Follow your day
Track meals, macros, water and activity alongside your goals.
DECISIONS THAT SHAPED THE PRODUCT
Turning AI into a usable workflow.
Photo selection, preview, analysis and results are separate states. Users can see when analysis is running and when a result is added to their diary. Nutrient estimates support everyday logging without presenting model output as a medical conclusion.
A product beyond a single screen.
Daily history, water and activity tracking, recipe creation and subscription access have distinct responsibilities. Firebase records and local caching support access to historical data. A separate web dashboard handles content and management alongside the mobile product.
THE RESULTING PRODUCT
From a feature to an everyday experience.
Live on iOS and Android in 11 languages, rated 4.4 on the App Store. Daily history, water and activity tracking, recipe generation and the subscription flow sit in one product, with content and management in a separate web panel. A performance pass prioritised startup tasks, removed unused libraries and split a 1,300-line subscription service into modules.
02 / Inside the product
FitZen screenshots
Open a screen to explore the details.06 images
03 / Scope & details
FitZen key features
- 01
Food recognition from a photo
- 02
Barcode scanning against an international product database
- 03
Firebase records, daily history and local caching
- 04
Macro and micro tracking, water and workout integration
Technology
INSIDE THE PRODUCT
The engineering behind FitZen
Systems we built and concrete problems we addressed in this project.
A controlled path from photo to analysis
Camera and gallery feed the same analysis flow. Permission denial, cancelled selection and invalid image data are handled as separate states, making each step understandable.
A retired model does not break the app
The model name is embedded in the mobile bundle; when the provider retires a version every released build gets a 404. An alias table in the Cloudflare Worker proxy maps retired names to a living model; exactly that happened for gpt-5.2-chat-latest on 3 September 2026 and no store update was needed.
Notifications and subscriptions on the server
Morning and evening notifications go out per language from Pub/Sub schedulers, and a smart scheduler running every ten minutes respects the user’s time zone. A RevenueCat webhook writes entitlements, a separate job reconciles expired premium, and dietitian session credits are consumed through onCall functions.
PART OF FITZEN · WEB
FitZen Web
The content management and analytics panel behind FitZen.
Challenge
As the app’s recipe and blog content grew, every update needed an app release. The content also had to be found by search engines and answer engines (GPTBot, ClaudeBot, Perplexity), which do not run JavaScript; a page that fetches data client-side is empty to them.
Approach
Content moved to Firestore and an admin panel; recipes, blog posts and quizzes ship without an app update. The site runs in 11 languages and everything renders on the server. Turkish is the default with no URL prefix (so indexed addresses stay intact); other languages use localised path segments such as /de/rezepte/…, and a request with the wrong language’s slug is 301-redirected to the canonical address. Recipe pages carry a self-contained, number-bearing "quick answer" block; FAQs are <details> in HTML; schemas sit under one @graph linked by @id; per-language RSS, an hreflang sitemap and llms.txt.
Result
fitzenweb.com is live; the production build generates about 2,000 static pages. Content updates stopped being a developer task and no longer wait on store review; user activity and analytics are read in the same panel.
Web
FitZen Web screenshots
Open a screen to explore the details.03 images
Web
FitZen Web key features
- 01
Content management panel
- 02
Recipe and blog management
- 03
Test and quiz creation
- 04
User activity and analytics
Technology
Two root layouts
app/(main)/layout.tsx serves Turkish and app/(intl)/[lang]/layout.tsx the other ten languages, because <html lang> and dir must be right and one root layout cannot read a dynamic segment. Slugs derive from the title in Latin-alphabet languages and from the ASCII form of the Turkish slug in Arabic, Japanese and Thai.
HTML written for answer engines
FAQ accordions are <details>, not JS; recipe pages carry a self-contained, number-bearing #quick-answer block; schemas sit under one @graph linked by @id (lib/schema.ts); llms.txt, per-language feed.xml and an hreflang sitemap.
About 2,000 static pages
The production build statically generates the recipe archive in 11 languages; api/revalidate refreshes affected pages after a panel edit. Bundle size is tracked with npm run analyze.
- 4,4 / 5
- App Store rating
- 32
- ratings
Your idea could be next.
What could we build for you?
From a first release to a growing platform. Let’s plan the design, development and launch together.
From Türkiye to the US, UK, Australia and everywhere in between.
ENGINEERING NOTES FROM THIS PROJECT
How to ship in-app subscriptions in React Native? Lessons from FitZen ↗FitZen: turning an AI result into an everyday workflow ↗