Skip to content
All projects
Health & Fitness2025AI-powered

Cal Care — AI Calorie Tracker

Point a camera at dinner. Get calories and macros back.

  • iOS
  • iPadOS
iOS 18+
Platform
SwiftUI
Built with
Vision + LLM
Model

About

A nutrition tracker that removes the part everybody quits over: the searching. Point the camera at a plate and an on-device Vision pass plus an LLM food-recognition model works out what is on it, estimates the portion, and writes the calories, protein, carbs and fat into the day. Around that scanner sits a complete tracking product — macro rings, weight logging, streaks, targets that adapt to whether you are cutting or building, and progress analytics that survive more than a week of use.

What shipped

  • AI food scanner: photo → identification → calorie and macro estimate
  • Automatic meal logging with editable portions and a searchable history
  • Daily calorie and macro targets derived from the user's stated goal
  • Weight tracking with trend charts and progress analytics
  • iOS 18+, SwiftUI, HealthKit-friendly data modelling, StoreKit subscriptions

How it was built

  1. Recognition pipeline

    On-device Vision pre-pass to isolate the plate, then an LLM food-recognition model for identification and portion estimation.

  2. Tracking product

    Macro rings, editable portions, searchable history and weight logging built on a HealthKit-friendly data model.

  3. Targets and analytics

    Goal-driven daily targets, streaks and trend charts, so the number on screen means something after week two.

  4. Monetisation and ship

    StoreKit subscriptions, entitlement handling and App Store release.

Challenges & solutions

  • ProblemManual food-database searching is where calorie trackers lose their users.

    SolutionMade the camera the primary input. Identification and portion estimation happen automatically; typing is the fallback, not the default.

  • ProblemA model estimate is a guess, and a wrong guess that cannot be fixed destroys trust in the whole log.

    SolutionEvery logged meal stays editable — portions, items and macros — so a bad estimate is a two-tap correction rather than a reason to stop logging.

Built with

  • SwiftUI
  • AI Food Scanner
  • Vision
  • Nutrition AI
  • StoreKit