Galaxy Embodied AI Agent
Samsung Research America
Embodied AI agent experiences for Samsung's XR platform, built on a definition framework and the interaction models that made an emerging industry concept designable.
ROLE
UX Design Intern, MPS IMEX (Immersive Experience) Lab
2 UX designers, 1 PM, 2 engineers
12 weeks
Samsung Android XR, Galaxy native apps
One scenario, three apps, no menus.
Watching a show → message arrives → press the physical button → agent summarizes it aloud → gaze to place the call → live transcription while they talk → "summarize the transcription in Notes" → shared with a colleague.
Three Galaxy apps, one request chain, no menu hunting. I designed eight streamlined use cases and shipped these three, because they chain into a single story and cover six agent features.
The agent's body came from Galaxy AI's own mark.
A four-point star derived from the Galaxy AI logo, given eyes for facial states and limbs for gestures. Simple enough to prototype in days, expressive enough to show what the agent was doing, and labeled a directional placeholder on purpose: real character design would have turned every review into a debate about the face instead of the interaction.
Two goals aimed at an object nobody had defined
Two goals: a business goal from Samsung HQ and a UX goal from the IMEX Lab. I surveyed how the field defined the object they both pointed at, across industry and academia, software and hardware, looking for something solid enough to design against rather than a term that only sounded advanced.
Eight traits became six capabilities of the AI agent
The field kept repeating eight traits, ranked by how often they appeared. I ran them through three filters, then changed the question from "what is an embodied AI agent" to "what would I have to draw to prove it." What came out was one operational definition, expressed as six capabilities, each with an interaction model that proves it.
Three use cases to build the PoC
Making the three cases real meant carrying Galaxy's OneUI design system onto a platform it was never drawn for. I migrated the patterns and built the XR components the prototype needed, panels, head-bar, dock, and the agent itself, then the team prototyped it.
Six features carried the demo. It went to US and Korea HQ as one continuous run rather than a feature tour, because the argument was that the capabilities work together or not at all.
The outcome
Impact & Feedback
Direction validated
Samsung Korea HQ confirmed the research and the capability framework matched the trajectory they were already on.
Scales past XR
Voice and tone carry embodiment on smart glasses, multimodal in smart home IoT, contextual presence on phone and tablet.
