Case study · Client name withheld

AI Calorie Tracker App

A native iPhone app that turns a photo, description, barcode or nutrition label into calorie and macro estimates

An AI-assisted calorie and nutrition tracker for iPhone. The client needed meal logging that turns a photo, a typed description, a barcode or a nutrition label into calorie and macro estimates. Our team built the native iOS app and the back-end services and AI endpoints that pass food analysis to third-party AI models.

Recreated iPhone result screen for an example chicken and rice bowl, 420 g, with an AI estimate of 540 calories, 38 g protein, 63 g carbs and 15 g fat, beside a daily goal card showing 930 kcal left; the food is a drawn illustration
Screens recreated with neutral naming and example data.
Client
Confidential client
Platforms
iOS
Our scope
Native iOS app · Back-end services and AI endpoints · Keeping the app responsive
Period
2025 to 2026

The product

What it does, and for whom

Food analysis takes time: an image has to be prepared, sent, analysed and stored. The challenge was keeping the interface responsive through all of that, and making sure a retry did not repeat an analysis.

Scope: The AI models themselves are third-party: there is no on-device or proprietary model, and our endpoints pass requests to them. Calorie and macro figures in the app are estimates.

Our contribution

What our team built

  • Native iOS app

    The iPhone app in Swift and SwiftUI with UIKit integrations, Combine, Swift concurrency and Core Data, including Apple Health sync through HealthKit and subscriptions with trials and restoration through StoreKit 2.

  • Back-end services and AI endpoints

    HTTPS endpoints for food-photo, nutrition-label and text analysis, nutrition plans, corrections and workout analysis, with third-party AI models behind them. Firebase handles sign-in, cloud data, images, messaging and remote configuration.

  • Keeping the app responsive

    Image preparation, AI requests and storage run off the main thread. Pending scans are saved with request IDs, so retries avoid repeating an analysis, and stored nutrition is reused for matching foods.

Technologies: Swift · SwiftUI · UIKit · Combine · Swift concurrency · Core Data · Firebase Auth · Firestore · Firebase Storage · Firebase Cloud Messaging · Firebase Remote Config · HealthKit · StoreKit 2

User flow

From plate to progress

Four screens in the order a user meets them.

Recreated screens of the app in order, left to right: capture a meal; see the estimate; today's goals and log; weekly progress
In the order of the steps below. Numbered markers match the steps below.
  1. Log the meal

    Take a photo, describe it, scan a barcode or capture a nutrition label.

  2. Check the estimate

    Calories, macros and nutrient notes come back, and can be corrected.

  3. See the day

    The meal joins today's log, counted against personal goals.

  4. Follow progress

    Calories, water and workouts build into a view over time.

The product in use

Inside the app

  1. Four ways to log a meal

    A meal can start as a photo, a typed description, a barcode or a nutrition label.

    Recreated capture screen framing an illustrated plate of food, with an enlarged view of its controls: Photo, Describe, Barcode and Label, and a shutter button
    Choosing how to log. Close-ups on phones: tap any image for the full screen.
  2. A result the user can correct

    The analysis returns calories and macros for the portion, plus short nutrient notes. If it looks wrong, the user can correct the result.

    Recreated result screen for an example chicken and rice bowl, with an enlarged card labelled AI estimate showing 540 calories, 38 g protein, 63 g carbs and 15 g fat; the phone also shows nutrient notes and Fix result and Done buttons
    Estimate for one meal. Close-ups on phones: tap any image for the full screen.
  3. The day against personal goals

    Today's screen tracks calories, macros and water against the user's own goals. An analysis still running appears in the log with its progress.

    Recreated Today screen with an enlarged goal card showing 930 kcal left and example protein, carbs and fat against goals; the phone also shows water, an analysis in progress at 30% and three logged meals
    Today's goals and log. Close-ups on phones: tap any image for the full screen.
  4. Progress and workouts

    Progress over time is shown against the same goals. Workouts are logged in the app and sync with Apple Health.

    Recreated Progress screen with enlarged panels: a weekly calorie bar chart against a 2,000 goal marked Example data, weekly water bars, and a logged workout synced with Apple Health
    Progress, water and workouts. Close-ups on phones: tap any image for the full screen.

About these images: the screens are recreated for this case study, with neutral naming in place of the client's brand. Meals, goals and figures are made-up examples, and the food is a simple illustration. The images follow the app's functions, but the layouts are illustrative and they are not screenshots of it.

Planning an app with AI at its core?

Tell us what the AI has to read or decide and what users do with the answer. We'll reply with the questions about models, cost and failed requests that shape the scope.