End-to-end experience architecture
I mapped the complete journey across login, onboarding, the main exhibition, content feed, character discovery, profile, chat, call and account settings.
Mystix AI
Mystix is an AI character platform designed as a continuous relationship—not a one-off chatbot. I planned the complete app experience and the AI systems that let characters discover, remember, speak and grow with each user.
mystix
Discover · Remember · Relate
CompanyMystix AI
RoleProduct OwnerUX Product DesignerAI Service Planner
DeliverablesEnd-to-end Product PlanUX/UI DesignAI System ArchitectureEpic · Story · Task Backlog
Project datePlanning, testing & iteration: Monthly cycles · Time to launch: 3 months
Context & problem
Character AI products can feel fragmented when discovery, conversation, memory and voice are designed as separate features. Mystix needed one coherent product model spanning the first login through an evolving long-term relationship.
“How might we make an AI character feel discoverable, expressive and genuinely continuous across every interaction?”
Product scope
One product.
Every touchpoint.
01Social login
02Onboarding
03My Page
04Main exhibition
05Content feed
06Character detail
07Character chat
08Character call
09Relationship graph
10Short & long-term memory
11LLM + knowledge RAG
12Voice & character generation
Product strategy & delivery
Persistent character model
Relationship states, short- and long-term memory, knowledge-based RAG and LLM orchestration were designed as one context system behind every conversation.
Sprint-ready product definition
The product was broken into epics, user stories and implementation tasks with acceptance criteria, then planned, tested and iterated in monthly cycles.
AI system blueprint
From a message to a relationship
Each surface was planned together with the context, intelligence and expression layers required to make the character feel consistent.
Relationship experience
Discovery, feed, profile, chat, call and settings form one continuous relationship journey.
Character logic & response
The LLM connects persona, conversation policy and relationship state to keep responses coherent.
Memory & knowledge
Short-term context, long-term memory and knowledge-based RAG give each character continuity and grounding.
Voice & character creation
ElevenLabs voice synthesis and Seedance-based generation extend the character beyond text.
Delivery model
Epic → story → task → launch
Monthly planning, testing and iteration kept product intent, UX decisions and AI dependencies connected from the first backlog through launch.
Epic
Defined the major product outcomes and end-to-end experience boundaries.
Story
Translated each outcome into user-valued flows with clear acceptance criteria.
Task
Connected screens, policies, content and AI integration work to sprint execution.
Launch
Repeated planning, testing and iteration in monthly cycles, then launched in three months.
Outcomes
A complete AI character product, launched in three months
Mystix connected a consumer-facing mobile experience with the systems behind believable characters: LLM orchestration, relationship progression, layered memory, knowledge RAG, ElevenLabs voice synthesis and Seedance-based character creation.
End-to-end product experience
Designed as one continuous journey
01
Enter with confidence
Account access, privacy disclosure and age onboarding establish trust before the first character interaction.
02
Discover a living world
The exhibition, feed and search experience make characters feel active before a conversation begins.
03
Manage the relationship
Profile, language, announcements and account controls keep the service understandable and operable.
04
Know the character
A detailed identity, relationship status and character-authored moments build context beyond the chat room.
05
Talk, listen and stay safe
Text chat, conversation controls and a live voice call turn the character relationship into a multimodal experience.
Impact