The Inspiration

It started at 2 am during exam season. We were students. Burnt out, sleep-deprived, running on deadline adrenaline and cold coffee. One of us noticed something strange: the moment a certain song came on shuffle, something in the body just… unclenched. Heart rate dropped. Shoulders dropped. Breath slowed.

We'd all experienced it. That one song that always works. But we had no idea why. And we had no tool that could prove it was real, let alone use it intentionally.

Mental health for students is invisible infrastructure. We talk about it, but we don't measure it. We don't see it. We don't act on it fast enough. By the time someone reaches crisis, weeks of quiet distress have already passed unnoticed.

That gap, between what your body knows and what you can act on- became our problem to solve.

"What if your music app knew you were stressed before you did, and responded before you even thought to ask?"

Challenges We Hit

Apple HealthKit wall HealthKit only runs on native iOS. We had to build a convincing biometric simulator that generated physiologically realistic data, a sine-wave heart rate with natural HRV noise, correlated stress indices, and 28-day cycle phase modelling. Real-time sync without chaos Keeping WebSocket data, Zustand state, chart history, and music recommendations in sync — without prop-drilling or stale closures- required careful architecture. Reconnection logic with exponential backoff took three rewrites. Translating science into UX The research was dense. Turning "cardiac entrainment via tempo synchronisation" into something a stressed student could understand in five seconds — that was the real design challenge. Every label had to earn its place. Spotify API rate limits The recommendations endpoint had strict constraints on seed genres and audio feature targeting. We built a mock song library as a fallback so the app works fully in demo mode, zero API keys required.

The gap between data and action is a design problem.

We learned that most wellness apps collect data but don't close the loop. They show you your heart rate. They don't respond to it. The intervention, the thing that actually helps, is always one more tap away. We built MoodMate to close that loop automatically.

We also learned that building with AI changes what's possible in a hackathon. Not because it writes code faster (though it does), but because it lets you think at a higher level. We spent more time on the problem and the science than on boilerplate. That's a different kind of leverage.

And we learned that biometric data is intimate. How you visualise it, label it, and talk about it matters enormously. The difference between a dashboard that feels clinical and one that feels caring is about 12 font choices and one colour palette decision.

What We're Proud Of

The mood engine, a weighted composite of heart rate, HRV, sleep, and cycle phase, actually mirrors the research paper's KNN classification model. It's not fake science. It's a real implementation. The dashboard works and feels alive. The BPM ring pulses at your actual heart rate. The mood blob breathes. The transition from "high stress" to "calm" takes exactly 2 seconds and feels earned. Zero API keys required. Every feature, biometrics, mood classification, music recommendations, charts, runs in full demo mode out of the box. A judge can run this in 60 seconds. We built something we'd actually use. That's the bar that matters.

Next Steps

The companion iOS app is the unlock. The web platform is the proof of concept. The real version streams live HealthKit data from Apple Watch, and that changes everything. Real HRV. Real sleep analysis. Real light exposure. Real menstrual cycle tracking. All feeding a recommendation engine that gets smarter the longer you use it.

1: Native iOS companion app with full HealthKit integration — live heart rate, HRV, sleep analysis, and cycle data streaming to the web platform in real time. 2: Feedback loop ML model — when you rate a song and report feeling better, that signal trains the recommendation engine for your physiology specifically. 3: Therapist/coach dashboard — aggregate anonymised mood patterns to give mental health professionals a longitudinal view of student wellbeing at scale. 4: University partnerships — pilot with student wellness programmes where the need is highest and the evidence base is clearest. 5: Expand beyond music — the mood engine can trigger breathing exercises, movement prompts, or rest reminders. Music is just the first intervention.