The domain where shipping one small app teaches more than reading the whole course — because the constraints here (process death, offline, store review) are ones you only internalise by hitting them. And both stores let you publish for a one-off or small fee.
Difficulty: 🟢 starter (hours–days) · 🟡 intermediate (a week or two) · 🔴 ambitious / portfolio-grade. ⭐ = highest signal.
The ladder
-
🟢 ⭐ Survive process death — build any two-screen app with a form, then turn on “Don’t keep activities” (Android) or kill the process from Xcode. Done when: you can background the app mid-form, kill the process, return, and lose nothing. Do this first — it’s an afternoon, and it teaches the constraint that shapes everything else → the lifecycle.
-
🟢 A list-detail app over a public API — the “hello world” of mobile, done properly: loading, error, and empty states as one sealed type, pull-to-refresh, and a lazy list with stable keys. Done when: all four states render correctly and the list scrolls 1,000 items smoothly.
-
🟢 Break your own network — take the app above and test it with Network Link Conditioner on 3G, with airplane mode toggled mid-request, and behind a captive portal. Done when: none of the three produces a crash, an infinite spinner, or a lie. This is the most commonly untested path in mobile and it’s the daily experience of a large share of users → networking.
-
🟡 ⭐ An offline-first app — local database as the source of truth, UI observes it, sync layer updates it, mutations queued in an outbox table. Done when: you can create, edit and delete with airplane mode on, then watch it all sync when you reconnect — and you’ve decided, deliberately, what happens when two devices conflict → offline-first.
-
🟡 Ship it to a store — actually publish. Privacy declarations, screenshots, a demo account in review notes, staged rollout. Done when: it’s live and a stranger can install it. The review process teaches things no note can, and rejections are the useful part.
-
🟡 Add push, end to end — server → FCM/APNs → device → deep link into the right screen from a cold start, with auth restored first. Done when: tapping a notification with the app fully closed lands the user on the correct screen, logged in → navigation.
-
🟡 Profile and fix a real regression — deliberately slow your own app (eager SDK init, a blocking network call at launch, an oversized image), measure the cold start and jank, then fix them. Done when: you have before/after numbers from Macrobenchmark or Instruments, not impressions → performance.
-
🟡 Make it accessible — full keyboard/switch navigation, labelled controls, and 200% text size without broken layouts. Done when: you complete the app’s main task using VoiceOver or TalkBack with the screen off.
-
🔴 ⭐ The same app, two ways — build one small app natively and in Flutter or React Native. Done when: you can name three things each made easy and three each made painful. This is the only way to have an informed opinion on the native-vs-cross-platform question, and it’s the answer interviewers actually respect.
-
🔴 Something that uses the hardware — camera with real-time processing, background location with a geofence, BLE to a device you built → hardware. Done when: it works on a real device, in real conditions, with permissions denied at least once.
If you only do one
Ship a small app to a store. Not because the app matters, but because the pipeline does — signing, privacy declarations, review, rejection, staged rollout, and a crash report from a device you’ve never seen. You cannot simulate that, and it’s the single clearest signal in mobile hiring.
Related
- the mobile course · interview bank
- frameworks — pick your stack
- Project Ideas — the vault-wide index
Software engineering
Infrastructure & security
Computer science
Hardware & engineering