UpDish
An iPhone app that looks at a photo of a home-cooked meal and shows how balanced it is, using Indonesia’s Isi Piringku guideline.
- My role
- Feedback and on-device AI developer
- Timeline
- Jun to Aug 2026
- Team
- Team of 5
- Platform
- iOS app


What I built
- Developed the meal evaluation and feedback screen in SwiftUI, scoring each plate across 4 food groups
- Generated on-device recommendations with Foundation Models and Translation, using 6 Indonesian portion units
- Cut the on-device prompt by 52% (844 to 405 tokens), which removed context-window failures across 12 on-device runs
- Engineered a 146-keyword offline food classifier that passed 26 of 26 edge cases, and cached results in SwiftData
Built with
- Swift
- SwiftUI
- SwiftData
- Foundation Models
- Translation
At the showcase
Team
With Andra Rachmantara, Andrew Wallace, Evelin Alim Natadjaja and Sherly Dinata Oey.