A personalized, voice-forward travel planner that helps multimodal commuters get around Seattle in a more time-efficient way.
Transportation and the growth of a future smart city go hand in hand — and Seattle, adding roughly 120K new residents with most commuters already mixing modes, was the test bed. As a two-person Master's capstone advised by Amazon, teammate Sophia Li and I set out to design a travel system for commuters who mix driving, biking, walking, and transit — and constantly hit friction doing it.
I led the research: recruiting 12 multimodal commuters from over 100 screener responses, running 7 semi-structured interviews, conducting a 150-minute "fly on the wall" observation during U-District morning rush hour, and running a diary study to capture real-life transit behavior — plus expert interviews with the Seattle Department of Transportation and UW to understand the policy layer.
Transit apps limit mobility behavior by only surfacing a narrow set of alternatives.
Lack of multimodal integration — technical, physical, financial — blocks new mobility habits.
Multimodal transit has to replicate the freedom and independence private cars provide.
From a participatory design workshop I facilitated, we ran Crazy 8's to generate 150 concepts, narrowed with a decision matrix against three design principles — seamless, personalized, durable — and landed on a concept combining transit accommodation, real-time info, and voice-driven recommendations.
Prototype testing surfaced real friction: a dark color palette caused usability issues, and 4 of 7 participants couldn't complete tasks using voice commands alone. That drove the dedicated voice UX pass I owned — designing for trip planning at home ("Puente" wake phrase or a tap) and quick in-transit customization, with explicit corrective feedback when the assistant misunderstood intent.
Wake phrase at home, tap in transit — with corrective feedback when intent is misread.
The research and insights were shared into the city's broader Smart Seattle conversation. My biggest lesson: designing a natural, seamless voice conversation takes as much craft as designing the UI around it — sometimes more.
Live modelPowered by GPT-4o Grounded in Sayena's case studies, not a canned bot. Billed per question — so make them good ones.