← All work
01 — Sleep Airline

Sleep, reimagined as a flight.

A bedside device that turns falling asleep into a “takeoff ritual” — GenAI captain's voice, illustrated cityscapes, and stories instead of sleep scores.

HCIResearch through DesignGenAIRaspberry PiUI/UX
My role
Solo — research, design & full build
Timeline
2025 Q4 – present
Build
Raspberry Pi · Firebase · GenAI API · Cursor
Status
ACM DIS 2026 WIP submitted · TAI-CHI Poster Submitted
Sleep Airline bedside device
Sleep Airline problem and design strategy
Portfolio slide · problem space & design strategy
The problem

When rest becomes a performance.

Sleep trackers turn rest into a score. Fixating on those numbers can backfire into “orthosomnia” — anxiety caused by the pursuit of perfect sleep data (Baron et al., 2017).

At the same time, sleep science focuses on mid-sleep, leaving the moment of falling asleep — the onset — largely undesigned. As HCI shifts toward experience-centered design, narrative design for this transition is still an open gap.

Design metaphor diagram
Design metaphor — one sleep session becomes one flight; hours of sleep map to flight distance and a destination.
The concept

One sleep = one flight.

Set your hours with a knob and the device plots a route: hours × 900 km/h = a destination. Nine hours lands you in Paris.

  • Distance over duration — waking up becomes a new story, not a test grade.
  • Narrative over data — GenAI writes a personalized captain's announcement for every session.
  • Minimal tracking — duration only, so there's no sleep-score pressure.
Interaction & UI

Designed for grogginess.

The two moments around sleep are cognitively fragile, so the interface splits in two:

  • The device — zero-reading, physical knob & button only; the screen lights only for takeoff and landing, then auto-dims.
  • The app — richer reflection delivered only once you're fully awake: ticket collection, on-time stats, duration trends.
Device and app interface decisions
Two states, two interfaces — a zero-reading bedside device for grogginess; a reflective app for when you're awake.
End to end user journey
End-to-end journey — set duration, confirm flight, sleep, cruise, land. GenAI rewrites the arrival story to your actual wake time.
The journey

A story that adapts to your morning.

Pull the shutter to depart; close it to sleep. Each morning, GenAI adapts the arrival narrative to when you actually woke up:

  • Woke early → “Wormhole shortcut — arrived ahead of schedule.”
  • On time → “Like a Swiss watch, landing perfectly on time.”
  • Slept in → “Severe turbulence — rerouting toward Türkiye.”
System

Hardware to cloud, built solo.

Using AI-assisted development (Cursor), I built a deployable, high-fidelity prototype end to end:

  • Hardware — Raspberry Pi with rotary knob, button, LCD and speaker, hand-soldered.
  • Software — GenAI API for narrative + cityscape generation; flight logic; Firebase sync.
  • Companion — a web app for flight logs and trajectory history.
System architecture diagram
Hardware-to-cloud architecture — Raspberry Pi → GenAI + flight logic → Firebase → companion web app.
Field deployment findings
Field deployment findings — behavioral shifts, meaning reconstruction, and unexpected social journeys across time zones.
What we learned

Behavioral shifts & a social surprise.

A 7-day in-home deployment with 3 users (the device as their primary alarm), with insights triangulated from daily LINE logs, interviews and review sessions:

  • Inviting — a retired officer reported natural drowsiness after the takeoff ritual.
  • Actionable — a night-shift nurse calculated bedtime backwards “to reach the destination,” sleeping earlier.
  • Shareable — one participant put down the phone to “protect the plot”; another created shared journeys with their sibling across time zones.

That social signal now drives V2: co-flying — shared flights, relays and layover-style sleep connections.

← All work