A trilingual product designer with 4+ years of experience designing cross-cultural and AI-driven digital experiences.
더 생생한 프로토타입을 원하신다면 링크로 가주세요! seyeonchoi.com
AutoView
Team
1 PM, 1 Designer, 3 Engineers
Role
Sole Product Designer
Timeline
Mar-May 2026
Platform
Web, Mobile
Overview
While managing a platform, I identified a structural issue: significant internal resources were being wasted on hardcoding repetitive frontend components. To solve this, the project was launched as an in house productivity tool where planners and designers could simply define data specs to generate React code via AI within a minute.
Following highly positive internal feedback, management decided to pivot the tool into an open source product targeting global developers. As our audience expanded to an external developer ecosystem, my challenge was to completely redesign the dashboard core, introducing features like GitHub integration, API key management, and collaborative sharing.
Pick your language:
The Problem
The time it takes for the LLM to analyze the schema and generate the initial React code. While the generation itself is highly efficient at under one minute, the first output rarely matches the user's exact needs, making subsequent edits inevitable.
Furthermore, the frequency of build errors during manual code modifications. When non-technical users attempted to tweak minor details directly in the generated code, it resulted in a 35% syntax error rate, causing the system to crash and disrupting the workflow.
Lastly, the abandonment rate during the component customization phase. Because users were forced to use text-based coding to make even the slightest visual adjustments, those without a programming background faced a steep learning curve and frequently abandoned the tool.
Average UI Generation Time
Over 1-2 min
Code-Level Error Rate
~35% on subsequent manual edits
Drop-off Rate
Significantly higher for non-technical users
User Research
"It’s amazing that AI can generate code in under a minute. But when it comes to tweaking minor UI details like padding or a header title, I still have to dive back into the raw code. For a non-technical user, the barrier to entry remains just as high." - Jane, Designer
Through user interviews (3 PMs, 3 PDs), I discovered a crucial pain point: users do not expect AI to generate a perfect UI on the first try. The real value lies in the iterative process, allowing users to converse with the AI to refine and tweak subtle details easily.
Hypothesis
If we combine a live preview with a natural language chat interface alongside the code editor, even non technical users will be able to build UIs effortlessly. Furthermore, a hybrid interface that blends chat commands with direct visual manipulation will provide the optimal balance of flexibility and precision.
Hybrid Editing Experience
If we combine a live preview with a natural language chat interface alongside the code editor, even non technical users will be able to refine and tweak subtle details effortlessly without touching raw code.
Role-Based Safe Workflow
If we design a distinct read only state for external teammates to restrict editing capabilities, we can prevent accidental code alterations and drastically lower the friction and drop off rate during collaboration.
Workboard
Design + PRD
Prototyping
I wanted to validate the usability of this complex interactive dashboard through an actual working environment rather than relying on qualitative assumptions. Utilizing Claude for coding, I personally built a high fidelity interactive prototype where the live preview rendered in real time and responded to natural language chat commands.
The Solution
While general AI builders like Claude exist, they fall short when it comes to managing component assets at an organizational level and integrating seamlessly with production codebases. AutoView bridges this gap. By designing a project based dashboard and implementing a secure, read only sharing system, I elevated the user experience from a simple AI coding sandbox to a robust, production ready development workflow.
Hybrid Dashboard Interface
I designed an interface that allows users to direct major layout changes via the chat panel on the left while modifying specific component attributes directly through intuitive popups on the preview screen.
Contextual Error Handling
When a user enters an invalid TypeScript interface, the system does not simply crash. Instead, it provides a clear, actionable guide at the bottom, helping users debug and continue their workflow.
Role-Based Sharing for Safe Collaboration
To ensure a secure workflow when sharing project links, I designed a distinct view for external teammates. This read only state restricts editing capabilities, preventing accidental alterations while preserving interactive testing.
Impact
Although the product was fully functional, management ultimately chose not to launch it publicly due to strategic decisions surrounding intellectual property and technical security. Instead, it was integrated internally as a core development asset, where it is now actively used by our teams.
100% Secured
Successfully integrated as a core internal tool
Actively Used
Currently in active use across our engineering and product teams
Wrtn Productivity Features
At Wrtn, I worked on the AI platform, which has grown to over 5 million monthly active users. The following two case studies showcase some of the key product challenges I tackled while designing and launching these experiences.
Pick your language:
Case Study #1
Beyond Fragmented Workflows: UI/UX Design for a Multi-Modal AI Summarizer
Team
1 PM, 1 Designer, 3 Engineers
Role
Sole Product Designer
Timeline
Aug - Sep 2025
Platform
Web, iOS, Android
Overview
Designed an integrated productivity service that lets users easily collect and summarize scattered information—such as YouTube videos, documents, and web articles—in one workspace, while allowing them to freely edit and repurpose the content through natural conversations with a friendly AI assistant.
In this project, as a sole product designer, I structured a cohesive information flow for various input types and designed the interaction model that guides users to refine their results through intuitive conversation.
The Problem
Clearly defined the structural limitations and psychological barriers users face when consuming long-form video and document data into three card components.
Tool Fragmentation
Users face severe friction because they have to switch between entirely separate sites and tools depending on whether they are watching a YouTube video, reviewing a PDF report, or reading a news article.
Cognitive Overwhelm
Users turn to summarizers to escape heavy reading. However, when the AI responds with another dense wall of text lacking hierarchy, it ironically triggers the exact same cognitive fatigue they tried to avoid.
Prompt Anxiety
Standard AI chat boxes feel rigid and mechanical. When users want to subtly tweak a summary, they often experience 'prompt anxiety,' struggling to figure out the exact robotic commands needed to get the right output.
Tool Fragmentation
Users face severe friction because they have to switch between entirely separate sites and tools depending on whether they are watching a YouTube video, reviewing a PDF report, or reading a news article.
Cognitive Overwhelm
Users turn to summarizers to escape heavy reading. However, when the AI responds with another dense wall of text lacking hierarchy, it ironically triggers the exact same cognitive fatigue they tried to avoid.
Prompt Anxiety
Standard AI chat boxes feel rigid and mechanical. When users want to subtly tweak a summary, they often experience 'prompt anxiety,' struggling to figure out the exact robotic commands needed to get the right output.
User Research
I analyzed real user voices to capture their frustration with rigid prompt interfaces and unorganized text layout, transforming these user paint points into concrete design opportunities.
"The AI summary is fine, but it's still so long and rigid that I end up highlighting and making my own notes anyway. It defeats the whole purpose of using a shortcut tool."
Users do not just need a shorter character count; they need immediate, actionable visual structures they can use right away.
"I never know what to type in the prompt box to make the tone softer or change it into a tweet. I'm tired of searching for prompt guide cheat sheets every single time."
We needed to shift the command-line paradigm into fluid conversational interactions and intuitive preset shortcuts to restore the user's sense of control.
Insights
This section highlights how the core insights discovered during user research were directly translated into key product features.
Unified Input Hub
To prevent users from dropping off across different tools, it was critical to create a single, unified interface that could absorb YouTube, Documents, Websites, and Text all in one place.
Visual Hierarchy Over Raw Text
Users want screens where the layout itself has already chunked and categorized the data—like timelines or bullet highlights—rather than being faced with plain, dense paragraphs.
Empathetic Agent
Instead of forcing users to stare at an empty input box and stress over commands, the interface needs an assistant with an approachable persona to bridge the communication gap.
Prototyping
The Solution
I designed Image Studio as a standalone product surface and unified image generation, transformation, and library management into a single workflow.
4-Format Integrated Input Hub
Reflected a 4-Format Integrated Input Tab Design using a clean tab layout, enabling users to paste and process any content format instantly within one workspace.
Cognitive Multi-Preset Layout
Placed a Multi-Preset Output Layout at the top of the results area to immediately relieve reading fatigue through visual chunking.
Conversational AI Character Panel
Integrated a warm, character-driven 'Roi' Interaction Panel on the right or when users request revision, guiding users to co-create and refine text through zero-pressure conversations.
Impact
Integrated Input Architecture successfully reduced the typical user drop-off rate during content aggregation by eliminating cross-platform navigation.
The visual layout structure utilizing Multi-Preset Tabs significantly enhanced data scannability, accelerating information digestion.
Conversational interactive refinement through 'Roi' minimized prompt errors and drastically shortened the path to generating finalized summaries.
~4x Faster
Task Completion Speed
-35%
Drop-off During AI Processing
Reflection
Bridging Technical Capabilities with Generatve UX
I realized that no matter how powerful an AI engine is, it fails to deliver a great experience if the outputs lack visual hierarchy or feel intimidating to control. This project reinforced that a product designer’s ultimate role is to humanize complex technical capabilities into intuitive, context-aware user flows.
The Importance of Empowering the User
Interfaces that force users to passively accept one-way AI outputs quickly lead to disengagement. This project was a valuable opportunity to learn how critical it is to grant users a sense of agency—allowing them to feel in control as they easily customize and co-create with the AI via preset shortcuts and organic conversations.
Case Study #2
Revolutionizing Generative Text UX Through Advanced Sentence Control Structures
Team
1 PM, 1 Designer, 3 Engineers
Role
Sole Product Designer
Timeline
Sep-Oct 2025
Platform
Web, iOS, Android
Overview
Text Humanizer is a feature that transforms AI-generated text into more natural, human-sounding writing while preserving its original meaning.
As generative AI becomes a common tool for drafting content, many users still spend significant time editing outputs due to repetitive phrasing, overly polished sentence structures, and a lack of personal voice.
Text Humanizer helps users refine AI-generated content by reducing recognizable AI patterns and adapting the writing style to better fit their intended context.
The Problem
While generative AI significantly improved writing speed, many users struggled to use the output without additional editing.
The writing felt overly polished and robotic
Similar phrases and transitions appeared repeatedly
User Research
Interviews with students and working professionals revealed that AI was primarily used to generate first drafts.
"Similar phrases and transitions appeared repeatedly"
"I can't trust the process of AI"
Participants consistently emphasized one requirement: "Make it sound like a real person wrote it." Also, users who had experience with paraphrasing tools often expressed concerns about losing meaning during rewrites.
User Research
The research revealed that users were not primarily trying to bypass AI detection.
Instead, they wanted:
Natural-sounding writing
Preserved
meaning
Transparency in what changed
Rather than rewriting everything, Text Humanizer needed to act as an intelligent editing layer that improves readability while respecting the author's original intent.
The Solution
Make changes visible
Instead of showing only the final output, we highlighted modified phrases so users could understand and review the changes.This increased transparency and helped users maintain confidence in the final result.
Adjustable rewrite strength
Different writing tasks required different levels of intervention. To support this, we introduced three rewrite modes: Light, Balanced, Strong
Users could choose how aggressively the text should be transformed depending on their goal.
Support different writing tones
Because natural writing varies by context, we introduced multiple tone options: Default, Friendly, Formal, Academic. This allowed users to adapt the same content to different audiences and situations.
Preserve important terms
Users expressed strong concerns about changing brand names, technical terminology, and key concepts.To address this, we introduced a keyword preservation feature that protects critical terms during rewriting.
Impact
Reduced friction in the writing workflow
Previously, users often moved between multiple tools to generate, edit, and finalize content By integrating Humanizer directly into the writing workflow, users could complete the entire process without leaving the product.
Increased trust through transparency
Making edits visible helped users better understand how their content changed. Rather than feeling replaced by AI, users felt more involved in the editing process.
AI Character Chat
At Wrtn, I worked on the AI platform, which has grown to over 5 million monthly active users, along with its AI character chat services. The following three case studies showcase some of the key product challenges I tackled while designing and launching these experiences.
Case Study #1
Redesigning the Visual Novel Experience Around the Limits of AI Generation
Team
1 PM, 1 Designer, 3 Engineers
Role
Sole Product Designer
Timeline
Nov-Dec 2025, Mar-Apr 2026
Platform
Web, iOS, Android
Overview
Kyarapu is an AI character chat service that Wrtn launched in Japan following the success of Crack in Korea. While exploring ways for Japanese users to enjoy more immersive AI-powered content, we saw an opportunity to create a visual novel experience where images and story progressed together, similar to a dating simulation game.
This idea was grounded in user behavior data. Kyarapu already included a feature that generated scene images for each chat message, and Japanese users were using this feature far more frequently than Korean users. I interpreted this as a strong signal that Japanese users placed greater value on experiences that combined storytelling with visuals, rather than text-only conversations.
Based on this insight, I proposed expanding image generation into a fully integrated content experience. The first version of the Visual Novel feature launched in December 2025. It generated a new image at every turn of the conversation, allowing users to experience an evolving story as they chatted.
Despite the team’s high expectations, the initial results fell short. We then turned to user data and qualitative feedback to understand what was disrupting immersion and identify how the experience could be improved.
The Problem
Post-launch analysis made it clear why the feature fell short of expectations.
Average scene generation time
10 - 15 secs
Average conversation length
~20 turns per user
Drop-off rate
Significantly higher than in the standard chat experience
Each image took more than 10 seconds to generate. Given that users typically engaged in conversations of around 20 turns, this resulted in several minutes of cumulative waiting time. The repeated delays disrupted the flow of the story and became a major barrier to immersion.
User Research
To better understand the issue, I interviewed six creators who had used the Visual Novel feature. Four of them identified the long image generation time as the biggest source of friction.
“I usually end up leaving because the wait feels too long.”
“It’s even more frustrating when I wait and the image still doesn’t turn out the way I wanted.”
Through interviews and behavioral analysis, I found that creators cared less about having AI generate images automatically and more about maintaining consistent visual quality that matched the world and scenes they had in mind. This preference was especially strong among highly active creators, who favored a more controlled workflow over generating a different image every time.
In other words, the core issue was not just generation speed. The combination of long wait times and inconsistent image quality made it difficult for creators to maintain control over the stories they wanted to tell.
Insights
Based on the research findings, I distilled the problem into three key challenges.
Beyond Speed Optimization
Reducing generation time depended on model and infrastructure improvements, so the design challenge was to find a UX solution beyond technical optimization.
Creator-Uploaded Assets
Allowing creators to upload their own assets enabled faster image delivery while maintaining consistent character and background quality.
Clear Usage Guidance
The key challenge was helping users understand when to use uploaded assets and when to rely on AI generation.
The Solution
I designed the system to prioritize creator-uploaded assets before falling back to AI generation. By allowing creators to upload image assets in advance, we were able to reduce generation time while maintaining consistent character and background visuals.
Designing Mode Selection in the Visual Novel Builder
I added a dropdown to the existing builder, allowing creators to choose between Asset Mode and AI Generation Mode based on their needs.
Structuring the Asset Library Experience
I designed the asset library to let creators upload and manage character expressions, poses, backgrounds, and scene images, giving them full control over building their own worlds.
Implementing a Drag-and-Drop Editing Experience
I designed a drag-and-drop interface that let creators easily place and organize uploaded assets by category, making large image libraries intuitive to manage.
Impact
Scenes built with uploaded assets loaded in under 5 seconds, reducing perceived wait time by 2 to 3× compared to the previous 10–15 second generation time.
Reusing uploaded assets reduced token costs and improved operational efficiency.
After launch, both user satisfaction and feature adoption increased.
~5sec
Loading Time
15+%
Cost Savings
Higher
User Satisfaction