Sanghyun Lee
← Back to work

Designing the assistant layer the rest of our AI products were built on

Ace

Uplimit was standing up AI experiences across the platform at once. I designed the Ace chat as the canonical persona pattern they all reused — one assistant across Spark creation, course authoring, live video conferencing, and the learner's view, instead of a different chatbot in each.

Ace — Designing the assistant layer the rest of our AI products were built on
Role
Product Designer (design + front-end)
Contribution
Systems · Interaction · Visual · Front-end
Platform
AI Assistant · Platform · Design System
Year
2025–2026

4

Surfaces built on the Ace pattern

1

Maintained code path, not four

2

Layers: persona + experience

Ace is your partner. Ace always knows everything about you. Ace is always there. And sometimes you have an ‘Ace session.’
Product lead — on the persona-vs-experience boundary

Context

Ace is Uplimit's AI assistant — an animated persona with a chat interface, reused across the learning platform. By early 2026 the same persona was showing up everywhere: generating Sparks, drafting course modules from a prompt and source material, auto-answering questions in live video conferencing, and riding along in the learner's view as a chatbot for any course-content question. That's the moment a platform either forms or fractures: one persona reused across surfaces, or a handful of subtly different “Aces.” Convergence wasn't a design preference — engineering wanted one code path for streaming, history, and voice, and coherence demanded one assistant that actually knows you.

The Ace persona — an animated character whose expressions track the agent's real state.
Ace in Spark creation — spinning up a Spark from a prompt.
Ace in authoring — generating course modules from a prompt and source material.
Ace in the learner's view — on hand to answer any question about the course.