Natural input
Let people describe the recipient in their own words, not just filters.
Case Study · Android · Kotlin
Gift Guide
A Kotlin/Jetpack Compose app that helps people find thoughtful gift ideas through guided prompts and personalized suggestions, instead of endless scrolling through generic lists.
Overview
The problem
Buying gifts is hard because it’s completely open-ended. People bounce between tabs, search “gift ideas,” and scroll through generic lists that rarely match the person they’re shopping for.
Gift Guide flips that by asking a few natural questions about the recipient and turning that into a small, focused set of suggestions that feel made for that person.
What I set out to do
Specific ideas
Make each suggestion clearly connected to the description they gave.
Real planning
Support saving and revisiting ideas instead of a one-time search.
Calm interface
Keep the UI quiet and readable, not noisy or “salesy.”
Core Experience
The app is built around three simple pieces: describing the person, getting ideas, and saving what resonates.
1
Describe the person
The entry screen asks a few short prompts:
- Who is this for?
- Occasion (birthday, graduation, “just because”…)
- Interests & personality
- Budget range
Inputs use flexible text areas and microcopy like “How would you describe them to a friend?” to keep things casual.
2
Review suggestions
After submitting, Gift Guide turns the description into a curated set of ideas. Each card shows:
- A concise title (e.g., “Custom recipe journal”)
- A short explanation tied to the description
- An optional price note or flexibility hint
The layout is a vertical list of cards—no clutter, clear spacing, easy to scan.
3
Save lists for later
Tapping “Save” adds the idea to a local list on the device. Users can:
- See all saved ideas across people
- Filter by person or occasion
- Mark items as purchased once they’re used
There’s no account or sync on purpose—the app behaves more like a planning notebook than a store.
Design Decisions
Two-tab structure
The app uses a simple “New Ideas / Saved” tab layout so people always know where live results and long-term planning live. No hidden drawers or deep menus.
Readable, quiet cards
Suggestions are designed like small index cards: clear title, one or two lines of explanation, and consistent spacing. The focus is the content, not decorative UI.
Helpful empty states
When there are no saved ideas yet, the app explains what will show up there and gives a direct call to action: “Create your first gift search.” This teaches the feature instead of leaving it blank.
Non-blocking feedback
Saving an item uses a subtle confirmation instead of a modal. The user stays in the flow of exploring ideas instead of being interrupted.
Architecture & Implementation
Under the hood
- Jetpack Compose for the UI, with state hoisted into ViewModels.
- ViewModel + state holders to manage inputs, loading, and suggestion results in a single source of truth.
- Local persistence (Room/DataStore) to store saved ideas, recipients, and simple metadata.
- A small recommendation endpoint that turns the recipient description into a list of ideas.
Why it’s structured this way
Keeping logic in ViewModels and using a single recommendation endpoint means the UI stays mostly declarative: each screen just renders its current state.
Local persistence is intentionally simple—no auth, no accounts—so the app is easy to install, try, and keep as a lightweight tool during holidays or big events.