1. Report detail
Felt unsafe
On your selected route — it counts toward the community flag in your preview.
Product design + engineering · 2025–2026
Fresh Greens
A navigation tool for Black drivers at different stages of a drive.
- Role
- Solo, design and engineering
- Scope
- Six interviews helped me frame three problems: Plan, Respond, and Trust.
- Outcome
- Working React Native prototype across 26+ screens.
- Timeline
- Sep 2025 – Jun 2026
- Tools
- Figma, Illustrator, Claude, React Native, Expo, TypeScript, Supabase
Car culture and South Jersey go hand in hand
I moved there from Brooklyn as a kid. Driving was the only practical way to get around, but I didn't know New Jersey like the back of my hand.
Whether I was dropping my mom at the bus station early or driving back from Cape May late, there were stretches that made me uncomfortable. The roads would start to break down a bit. Confederate flags would appear. Reception would become unreliable. All of the spooky pieces started falling into place.
I'd slow down, avoid backroads, and keep my wallet and phone close. Then I'd wonder what would happen if the car broke down right there. I didn't know whether that discomfort was mine alone, so I spoke with six Black drivers across the South, ages 19 to 34.
What I heard from six Black drivers
Qualitative interviews were new to me, but I tried to navigate them like everyday conversations. Unfortunately, I wasn't the only person getting the heebie-jeebies during a drive.
Drivers already relied on tools like Google Maps.
Plan
Drivers couldn't compare the conditions they cared about across routes before choosing. All six mentioned daylight, five mentioned road conditions, five mentioned police presence, and three mentioned wildlife.
Respond
Drivers talked about police encounters and keeping help close when something went wrong.
Trust
Five of six asked family or friends before trusting an unfamiliar place.
Six interviews can't represent every Black driver.
1. Compare route conditions before choosing
I knew Fresh Greens had some utility. Begrudgingly, my first idea was a Google Maps plug-in. I showed it to my thesis advisor. He was content, not impressed. The thesis was getting lost inside Google's framework. I was about a month from delivering it, so I kept the information architecture and started over visually.
I returned to the Green Book for the visual system. Its palette became a reference, not a claim that Fresh Greens is its digital successor. Source: Smithsonian National Museum of African American History and Culture.
On a trip from Chicago to rural Georgia, the driver needs to see where natural light fades, where artificial light picks up the slack, and what unknowns to account for. The driver isn't expecting perfection, but they are expecting the clarity and autonomy to choose a route that they can feel adequately prepared for.
The current build still labels the top option "Safest route," ahead of the research. It needs to explain the tradeoff and leave the choice with the driver.
Implemented route controls
Compare the conditions attached to each option.
1 of 3
10 hr 48 minSafest route with current conditions
- Arrival
- Arrive 7:10 PM
- Distance
- 708 mi
Low lightRoad
More daylight near departure
View the visual pivot and daylight reminder
First pass
I first put Fresh Greens inside Google Maps. The safety criteria were there, but Google's route still controlled the screen.
Standalone app
The standalone preview lets the driver compare daylight, route conditions, and trusted places before choosing.
Daylight came up in all six interviews
A daylight departure reminder.
I built the reminder around that pattern.
- 1. Pick a timeLeave before the light fadesSchedule
- 3. RemindFresh GreensnowTime to head out
Leaving now gives you more daylight.
What this showsRoute cards show some of the reasons behind a recommendation. Fresh Greens waits until someone taps Schedule before asking for notification access.
BoundaryThe prototype does not prove that a preferred route is safer. Its current Safest route label overstates the evidence, and the reminder has not been shown to change behavior.
2. Keep four support paths one tap away during a stressful moment
Worst comes to worst, a Black driver needs to be informed. A police encounter can be a lot to juggle. One tap opens four support paths, including the pulled-over flow.
Interactive case-study reconstruction
This shows one path through the prototype.
Toolkit
The toolkit starts with ‘What’s going on?’
Four paths stay behind one control.
Question
One question comes before the microphone request.
All three answers carry equal visual weight.
Choose an answer to continue.
Reassurance
Recording starts when microphone access is available.
Guidance comes next, with a trusted contact one step away.
Contact
Recording and trusted-contact actions stay visible.
The screen confirms that nothing has been sent.
This web reconstruction does not place a call, send a message, or share location.
I haven't tested this flow with drivers yet, let alone during a real encounter.
3. Show what influenced a route recommendation
People just need transparency into how the app is doing what it claims it does. Route chips and cards show public sources and community-report influence. The full score and its weights aren't exposed yet.
Reports stay on the device first. With Supabase configured, they enter moderation.
Current prototype limit:One report creates a scored zone and can affect route ranking. I haven't added corroboration weighting, visible contributor provenance, or route-level trust tiers yet.
Implemented route link
Show where a community report affected the preview.
2. Route preview
10 hr 42 min2 of 3
6 min faster
Community flagView the moderation path
Configured cloud
Queued
Investigation
- Source device
- Same-spot reports
- Nearby reports
- Duplicate IPs and devices
Decision
Reviewed, hidden, restored, or removed
What this showsReports stay on the device first. With Supabase configured, they can enter the moderation path.
BoundaryOne report can affect route ranking now. Corroboration from distinct contributors, visible provenance, and route-level trust tiers are not built yet.
What I built and what I still need to test
I used Figma to set the initial rules, then built them in code. A spacing problem became a shared theme rule.
Search broke first. My own address was sitting in Recent, so a places-only search looked functional. I added street addresses and separated Recent from live results.
The prototype still hasn't shown that a route is safer, its explanations earn trust, or its support flow holds up under stress.
Next, I'd learn from Black drivers 55 and older by watching where Fresh Greens fits, where it asks for too much, and what I need to change.
Built now
- Plan: route comparison, source cards, departure and refuel reminders
- Respond: four support paths, on-device recording when available, and user-controlled contact handoffs
- Trust: local-first reports and configured moderation
What remains
- Plan: broader route-quality testing before any safety claim
- Respond: real-device stress-state and failure-mode testing in configured builds
- Trust: route-level corroboration by distinct contributors, provenance, trust levels, and moderation transparency
