Implemented the initial bubble graph discussed during onboarding.
Improved it with significance-level labels.
Added categorical grouping on hover.
Improved cursor tracking for better interaction quality.
Feedback from the team
"very nice work Pin Zheng"
"love the tooltips and the categorical grouping on focus."
These messages matter because they show the quality of the work was
visible early, before the scope broadened into larger UX problems.
Problem Framing
I moved from solving a task to diagnosing the deeper product problem
Problems I identified
The existing mobile experience compressed and broke visually.
There was no strong central design library guiding decisions.
First-time users had weak product orientation.
The UI did not always clearly surface product value.
What I did with that
Reviewed the codebase and used ShadCN as a practical foundation.
Started building a lightweight design direction instead of isolated screens.
Treated onboarding and product education as part of the UX problem, not separate from it.
UX Direction
Design-system work and UX direction became a reusable product layer
What I explored
ShadCN-based design kit for spacing, colours, typography, and component consistency.
Mobile-first dashboard and graph views during the early redesign phase.
Desktop-first redesign after alignment on real user behaviour.
Chat as a core interaction surface rather than a side panel.
KPI cards, alerts, overlays, filters, and chat states for clearer flows.
Why it mattered
This work created a more coherent direction for how MetaLearner
should guide users, not only how it should look. It connected
visual consistency to product clarity and adoption.
Best supporting visuals for this slide are the mobile and desktop
design screenshots from the internship chat thread.
Onboarding and the User Guide became a real implementation track
What was implemented
Onboarding modal with swipe gestures and step navigation.
"Find out more" links into a User Guide.
User Guide hub page plus guide-detail pages.
4 guide categories: Dashboard Overview, Drill Down into the Details, Ask and Extract Your Data, and Settings.
Prompt-writing guidance and suggested-question guidance for first-time users.
Proof from the internship build
This frame was extracted from the WhatsApp demo video. Use it
together with the user-guide screenshots from the internship chat
thread for the strongest visual proof.
Why It Was Strong
The onboarding work solved adoption, not just presentation
Better orientation
The guide pages kept the left sidebar visible so users stayed
grounded in the product while learning it.
Lower interruption cost
"Find out more" links opened in a new tab so onboarding flow was
not broken.
Better prompt quality
The guide content included prompt-writing help, suggested-question
usage, and multi-turn conversation framing.
Validation
Lim’s feedback explicitly highlighted the value of the user guide
as a way for users to access detailed instructions.
"I like the flow, especially the addition of user guide which user could
look at detailed instructions on how to use the product."
Research Artifact
Real-time messaging research repo for future team use
It gave the team a concrete reference architecture for future
real-time collaboration or notification features without adding
immediate production scope.