A photo-op discovery feature, designed end-to-end in a five-day sprint, so people can find the best spots for photos near them.
A brief, and five days.
This one's a self-directed exercise — a UX prompt I ran to flex a single muscle: turning a brief into a credible, tested solution, fast.
GramCity is a photo-editing app. The brief: add a way for users to discover where to take great photos.
The user research here was provided as part of the brief — I didn't run the interviews. My work started at synthesis and ran through ideation, design, prototyping, and validation. The point of the exercise wasn't novel research; it was how quickly I can shape given inputs into a solution worth testing.
Five days, five pins.
A sprint trades breadth for speed — one focused pass from brief to validated prototype. Here's the route:
Cluster research, define user behavior, frame challenges.
Prioritize features, study rivals, ideate fast — by hand and with AI.
Storyboard the flow, build the HiFi screens.
Wire the screens into a clickable flow with Claude Design.
Test with five users, find friction, recommend fixes.
Synthesize the signal.
Starting from the provided interviews, I clustered the raw notes into themes using an affinity map — turning scattered quotes into patterns I could design against:
Two patterns ran through the clusters and shaped everything that followed:
I reframed those patterns as How Might We questions — the design problems I built the solution against:
Go wide, then choose.
Generate a lot, then narrow to the few features that actually carry the experience — and borrow patterns users already know.
Prioritized from the insights and the user flow. The core is the nearest-photo-ops task; search, saving, and filters are supporting scaffolding any real app needs — but not what a five-day sprint had to prove.
Apps that nail map + list discovery — patterns I could lean on.
Browse without losing your place.
Jump straight to the type you want.
Proof that a spot's worth the trip.
Eight directions in the time it'd take to hand-sketch two, using Claude Design to move through structural bets fast. From there, I sketched the chosen direction into a three-panel storyboard by hand, translating the fastest idea into the flow I'd actually build.
4 directions — Claude Design
Committed storyboard — hand sketched, map-first, with the list a tap away.
Sketches into screens.
I mapped the end-to-end flow first — so every screen had a clear next step — then built the sketches out in Figma. What's shown here is the three-screen spine from the storyboard — the rest of the flow (permissions, list view, reviews, photos) was built out in full there too, but this is the throughline that carries the story.
The critical path, start to finish — enable, browse, decide, get directions.
Leads with the payoff, so granting access feels worth it instead of intrusive.
Nearby photo-ops drop as pins — filter by type, or flip to a list, without losing the map.
The proof to commit: ratings with review counts, directions, and visitor photos.
Make it clickable.
The full user-screen flow became a clickable prototype in Claude Design — I fed it the finished HiFi screens and had it wire the taps, toggles, and transitions between them, so the interaction logic came from directing the tool, not hand-building each hotspot. Search and saving sit in the UI as the extras any real app needs — but the walkthrough stays on the one task that matters: find a spot, decide, go.
Put it in front of real users.
I tested the core task — finding a photo-op nearby and deciding whether it's worth the trip — with five participants, watching first impressions and how they moved through it.
Users moved through the task smoothly and responded well — the direction validated. It surfaced one refinement worth making.
Users wanted to enlarge other visitors' photos before making the trip — seeing the kind of shots people actually got at a spot is what makes you want to go shoot there yourself.
Fix: make the photo grid tappable — a full-screen, swipeable viewer so the vibe comes through.
Get in, find it, decide.
Five days in: a location-discovery feature that lets GramCity users explore photo-ops near them or in any city — with the proof to decide before they go.
Permission up front, with the payoff stated plainly.
One toggle flips between pins on a map and a scannable list — same spots, your way.
The proof to commit: ratings with review counts, directions, and visitor photos you can tap to enlarge.
What this
project taught me.
What this exercise really tested wasn't research — it was how fast I can turn a brief into something worth putting in front of users, and how much of that speed now comes from knowing when to hand a step to AI and when to do it myself.
Five days left no room to gold-plate. Every choice had to earn its place — and the work was clearer for it.
Leaning on Airbnb and Yelp conventions gave users familiarity they didn't have to spend effort learning.
Five users were enough to validate the direction and surface one real refinement. Validation doesn't need to be big to be useful.
Claude Design let me test more directions in the same five days. Picking which one to build still came down to me.
I'd run my own discovery round to pressure-test the provided research, build out the plan-ahead and saved-spots flows in full, test type-switching as a focused task, push Claude Design further on more complex interaction states, and re-run the photo-enlarge refinement to confirm it lands.
