Interactive Showrooms — End-to-End Design

Qwen Character is an LLM API — teams build their own character products on it, the way they build on Claude. But feeling the model meant configuring, running samples, and reading output alone — a loop that routinely stretched into hours, and most trial users left before the moment of value. So I redesigned its site into a storefront: a shift from documentation to proof.

How might we

Make model capabilities visible, testable, and trustworthy — within minutes?

The showroom strategy came from our PM — a familiar consumer-product play. My leverage was pushing it to its limit and owning the build: rather than improving the documentation, I designed 4 market-specific showrooms — companionship, psychotherapy, character cloning, IP licensing — that let users experience a working version of their own future product. On a 10-person team, every feature decision across all 4 rooms was mine, and all of it shipped on the live site. Users don't believe descriptions, so the first message had to prove the capability.

Three model strengths crammed into one chat window — none landed. So I split them across rooms: each makes one form of cognition visible — memory, analysis, or implementation — with one proof moment legible in 60 seconds, no explainer text.

Showroom → One proof → In-product behavior

Romance Room

Long-term memory

In experience

Character recalls conversation specifics across sessions

Astrology Room

Real-time memory updates

In experience

Live constellation profile updates mid-conversation

Therapy Room

Real-time analysis

In experience

Expert panel surfaces conversation themes as you chat

From here, each feature up close — the user psychology it's built on, and the model capability it makes visible.

I prototyped all four, but Moments Feed became the reusable core — not the flashiest, but the one direction that satisfied the most constraints at once. It proves two capabilities in a single surface (long-term memory and real-time reaction); it's the most cost-controllable for both model inference and engineering, unlike Alternate Universe's per-user live generation; it reuses the existing chat layout untouched; it lives in the same right-side drawerevery showroom already shares; and “a moments feed” is the most immediate mental model for users. So it became the template the other rooms were built from.

AI compressed the distance between strategy, visual direction, motion, and implementation. I used it to stress-test decisions, generate visual directions, prototype motion, and deliver production-adjacent interfaces that engineers could merge with minimal revision.

Inspired by Love and Deepspace, I used Wan, Kling, Dreamnia, and SeeDance for visual identity. Interactions built with Cursor and Claude Code — all in four weeks.

Live interactive prototypes

Open full page for best fidelity.

I delivered an end-to-end update to the Qwen Character admin — spanning API surfaces, Studio (Applications, Workflows, Knowledge Base, Characters), and the nested flows underneath. That included secondary screens teams rely on in production: empty and error states, and analytics views for operational data such as invocation metrics and call volume.

Design framework

View the live showrooms →