ProductGraduate capstone

MYND

A mental health app that reads behavioral signals and intervenes before a crisis, not after one.

The problem

More than one in five US adults live with mental illness. The barriers our five clinician interviews named weren’t clinical: culture, language, cost, stigma, time, convenience.

The gap isn’t treatment. It’s the distance between needing help and knowing help exists.

We ran a heuristic analysis of Daylio, Bearable, Garmin’s Body Battery and uMore. All four tracked. None intervened. That absence became the product.

What I did

I joined a startup with a mid-fidelity prototype, no cohesive look and no user testing.

I built the brand and a design system first, so later decisions had somewhere to resolve. Then I rebuilt the app around a single score, from sleep, HRV, steps, location, phone use and mood.

The anonymous survey surfaced coping mechanisms no interview did, including drinking, smoking and staying in bed all day. So the app couldn’t assume someone’s existing methods were healthy.

Two clinicians warned that a phone app contradicts their own advice to cut screen time. So interventions had to be rare and quiet.

That’s why the flow below starts on the lock screen, not in the app. Every step after it is ordered by what it costs you: someone you trust before a professional, a professional before an emergency line.

Walk the flow

Open the Figma prototype

1. The nudge arrives. A time-sensitive notification, not an app you have to remember to open. This one is a friend asking, which is the cheapest intervention there is.
2. One score. Sleep, screen time, steps and heart rate variability resolve into a single ring. A number you read in a second, not a dashboard you study.
3. The signals behind it. One tap from the ring. Mood, steps, sleep and HRV over time, so the score is auditable rather than a black box.
4. It asks before it acts. A low score is a hypothesis. The check-in validates it against how you actually feel before anything is recommended.
5. A low-risk intervention. Specific and finishable, then rated. The rating is what lets the app learn which mechanisms work for this person instead of guessing.
6. Or a person. Contacts filterable by what each is willing to talk about, so reaching out costs one tap and no explanation.
7. Escalation, by cost. Local support, then online, then in-app. Grouped so the cheapest option is the first one you see.
8. The last resort, last. 988 and emergency services are reachable in two taps but are never the first thing offered.