2 Months
Academic Project
Prototype
The design challenge: Professionals with urgent, specific questions face a frustrating gap in the market. Build a marketplace platform where professionals can book 15–60 minute paid consultations with specialists and walk away with actionable next steps.
My role: Responsible for all UX research synthesis, creative strategy, interaction design, and design system documentation.
Mapping the Full Journey
Two task flows: Clients searching for specialists & Completing the consultation session with AI assistance
Five Usability Testing Revealed Structural Friction
AI Recommendations Made Visible and Editable: Specialist selection lacked decision criteria: "Why is this the right one for me?" was unanswered
Systematic CTA Hierarchy: Global "Book session" CTA was contextually confusing on the home screen before any specialist was chosen
In-Call Timer Becomes a Proactive Session Manager
Evidence: "time notification could be more prompting and visible", "if mentor has availability, like a notification?"
The panel becomes reactive rather than static: a context-aware assistant that responds to session state.
Accessibility Documentation
Every colour combination appearing in the UI was tested against WCAG 2.1 using the Colour Contrast Analyser.
Suggested visual: The contrast grid table showing all 18 pairs with foreground/background colour swatches, ratio, and pass/fail badges, with the AI yellow row highlighted
Specialist Profile Restructured Around the Booking Decision
A structured information block: a quick stats row, a skill tag, the AI match rationale card, and a "recent session topics" strip showing anonymised examples of what this specialist actually covers in practice.
Building a Visual Language for "Credentialed Speed"
Yellow = AI
Every AI-generated element carries the
#FBFFAByellow with the✦sparkle icon, making the platform's intelligence layer immediately recognisable and transparently labeled.
Building an AI-assisted marketplace prototype taught me that the hardest design problems aren't about what's visible on screen; they're about what the user never finds. The preparation screen discoverability failure, the AI recommendation that didn't explain itself, the timer that no one noticed: all of these were features that existed but failed to do their job because of hierarchy, placement, and affordance decisions that looked fine in isolation.