Biscuit: AR virtual assistant.
Overview
BloomingPods is a speculative transit system that introduces autonomous feeder shuttles (pods) that pick students up from neighborhood streets and connect them to the city bus network.
To support this system, I designed an AR-based virtual assistant that consolidates scattered transit cues such as bus tracking, route confirmation, and stop awareness into a contextual spatial interface.
Instead of repeatedly checking apps or scanning the environment, students receive ambient guidance and real-time updates only when relevant.
The result is a context-aware commuting experience that reduces cognitive load while preserving user control during everyday transit.
Background
Extended independently from a group project
My contribution
All UI screens, prototypes, and AR system designs are original individual work.
My Role
UX designer, UI designer, UX researcher.
Tools
Figma, Figjam, Miro
Timeline
January 2026 - February 2026
One step ahead of your day
PROBLEM SPACE
IU students juggle 3+ apps
just to catch a bus and still miss it.
IU students juggle multiple bus apps, watch digital "stop requested" boards, and crane their necks to read unfamiliar streets — all while mentally managing class schedules and deadlines. When buses disappear from apps, arrive full without warning, or stop unexpectedly, the fragmented information landscape turns a routine commute into a stressful task.
We designed Biscuit, an AR-powered companion that consolidates scattered transit cues into a single, ambient spatial interface. Information surfaces only when relevant. The rest of the time, the system stays out of your way.
"When in a rush, checking SPOT (bus tracking app) actually slows me down."
— IU Graduate Student, Research Interview
RESEARCH & DISCOVERY
Finding the moments
where uncertainty becomes stress.
Our research goal was specific: to identify where and why uncertainty occurs during IU students' bus journeys, and locate the moments of cognitive stress most addressable through context-aware AR support. We combined field observation and interviews with bodystorming, a method that revealed interaction problems invisible on a 2D canvas.
01
Field Observation
Rode IU buses across multiple trips. Observed micro-behaviors: how riders look for seats, confirm stops, and judge whether their pace will get them there on time.
02
Interviews
Deep dives with IU students on commute pain points, trust levels with automated systems, and notification tolerance under academic stress.
03
Bodystorming
We physically simulated the bus journey using paper prototypes and uncovered a key issue: a centered navigation bar would block passengers’ faces in crowded conditions.
FROM RESEARCH TO DESIGN
Every insight has a design decision.
Students weren't confused about routes, they were overwhelmed by scattered reliability cues. Each finding below drove a specific design response.
Research findings
Design Decision
"Graduate students want to commute with their brain off."
Designed an ambient color-state system (translucent / blue / green) that communicates status without requiring reading — no text until the user taps.
Students couldn't see empty seats in the back of a crowded bus.
Floating AR markers identify vacant seats, even those hidden from boarding-line sightlines, and mark the correct pod on the curb.
Buses disappear from real-time tracking apps without explanation.
A persistent progress bar in the bottom-right corner with a one-tap delay reason and rerouting option which never silently disappearing.
Cold winters intensify pressure to time departures perfectly.
Pace-maker AR footprints guide walking speed so students leave at exactly the right moment, no mental calculation required.
Bodystorming: centered navigation overlays passengers' faces on a crowded bus.
Navigation bar repositioned to bottom-right corner.
BISCUIT - YOUR VIRTUAL ASSISTANT
Biscuit is the mobile onboarding and coordination layer for the BloomingPods system. It syncs with a student's class schedule, live bus data, fitness activity, and smart home — so the AR experience is personalized from the moment they step outside.
Welcome Screen
The app is introduced through a minimal, friendly splash — establishing Biscuit as an intelligent companion rather than just a transit tracker. The mixed-weight typography ("Biscuit") signals a playful but reliable personality.
Capabilities Overview
Rather than a settings list, Biscuit explains its value through four grounded use cases: class schedule syncing, live bus tracking, health nudges, and smart home prep. This mirrors how students already think about their day.
Permission & Trust
Each integration is opt-in with a plain-language explanation. Smart Home defaults to off — respecting user comfort levels. A privacy note ("your data stays on-device") directly addresses the biggest concern surfaced in research interviews.
Notification Alert
Customise the notification you want your assistant to send you. For lesser cognitve load choose only those that are necessary. Biscuit recommends which notification alerts to run.
AR SYSTEM DESIGN
Information in the periphery. Clarity on demand.
The AR layer is organized around two principles:
Progressive disclosure - shows detail only when requested &
Ambient status - uses color as language
The virtual assistant is a customizable AR cat and is an embodied guide that is present when helpful and invisible when not.
Translucent
System is idle. No active transit event. No attention required.
Blue — Update
Pod arrival incoming, transfer alert, or schedule change. Tap to expand.
Green — Wayfinding
Active navigation mode. AR footprints guide walking pace in real time.
Pace-maker Footprints
The AR takes a physical form of a cat that guides the user to their pod destination. AR footprints guide users to walk at exactly the right speed to reach the pod pickup spot on time. The assistant glows blue when pace is aligned; it dims when the user needs to slow down.
Floating Identifiers
AR markers float above empty seats, even those hidden in a crowded bus. Based on a direct field observation: students boarding a full bus couldn't see vacant seats in the back row. The identifiers are present also on ther curb where the pod is going to stop, and on top of the pod to identify you are getting on the right pod.
Progressive Status Bar
A minimal progress bar stays in the bottom-right corner throughout the ride. A red line signals delay — tap to expand for cause of delay, updated ETA, and rerouting options. Default view stays clean.
DESIGN OUTCOMES
The AR layer is designed around progressive disclosure and ambient status — color-coded states rather than text alerts, with detail panels available on tap. The virtual assistant (customizable avatar) acts as an embodied guide: visible when helpful, invisible when not.
Reduced Cognitive Load
Relevant info surfaces only at the right moment — departure timing, pod ETA, seat availability — eliminating manual app-switching and time calculations.
User Agency Preserved
Every VA suggestion can be overridden or dismissed. Students wanted help, not control. The system assists without automating judgment or removing decision-making.
Adaptive Over Time
The system learns from previous journeys and individual habits, gradually personalizing timing nudges and route suggestions without requiring manual input.
What I learned.
Bodystorming reveals what wireframes miss.
Physically acting out the bus journey showed us that a centered navigation bar would overlay passengers' faces in a crowded bus — a problem invisible on a 2D canvas. Moving it to the bottom-right corner was a decision that only emerged from the body.
Automate routine, not judgment.
Backed by Parasuraman et al.'s automation model, we focused on the repetitive: arrival time calculations, transfer syncing, pace guidance. High-stakes decisions (rerouting, canceling a trip) always stay with the user.
Privacy must be designed, not appended.
Integrating Canvas, location, and health data raised real concerns. The permission screen defaults Smart Home to off and reassures users their data stays on-device — but a full deployment would need deeper privacy architecture.
LIMITATION & NEXT STEPS
Honest constraints in the current scope.
-
Privacy Architecture: Privacy is one of the major concerns since our virtual assistant is collecting personal data from Canvas, calendars, and location tracking. We require some better controls and options so users know exactly what's being collected, from where its being collected and can opt out whenever required.
-
Real-Time API Reliability: Real-time data integration with traffic and weather apps requires APIs that might not always be reliable or accurate.
-
Safety in Adverse Conditions: We also haven't fully thought about the safety during the walks, especially in bad weather or at night. Weather variability in general is challenging. Heavy snow or rain could mess up pod routing and the physical attaching process between the pods and buses.
-
Spontaneous Trips: Our design doesn’t tackle what happens when pods or buses are full and in those cases users need to be rerouted. And finally we focused mainly on regular scheduled trips but haven't explored how the system would work for spontaneous out-of-schedule trips where there's no pattern data for the VA to learn from.
FINAL PROTOTYPE
UI Design - Light Mode
UI Design - Dark Mode
