AI assistant for product design

Project Overview

Kami is your AI co-pilot for product design, guiding you from research to design without tool overload, messy data, or creative burnout.

Project Type

Class Project (HCI for Interaction Design)

Timeline

Jan 2025 - May 2025

Role

Lead Product Designer, User Researcher, and Prototyper

Team

Clayton Lin, George Kim, Gabriena Feng

Problem

Designers often switch between too many tools that don’t work well together. This leads to wasted time, more mistakes, and makes it hard to create designs that are user-friendly, visually appealing, and usable.

Goal

Build a smart platform that lets designers handle everything in one place. This reduces tool switching, lowers errors, and helps create better designs for real users.

Key Features

Here are our final solutions for each core flow. They solve key pain points in product design and create a smooth, end-to-end workflow.

1

Research with Kami

Tell Kami the research topic, it will search, organize, and refine the information into clear, well-designed summaries.

2

Brainstorm with Kami

Tell Kami your goals or creative direction, and it will generate structured ideas, explore alternatives, and guide you toward actionable concepts.

3

Design with Kami

Turn wireframes into polished hi-fi mockups with clarity and consistency.

Tell Kami your preferred design style, and it will adapt the UI style while protecting the integrity of the experience.

Generate new interface designs based on your existing visual style.

4

Prototype with Kami

Turn your design into a working app

Secondary Research

We started with market research, we focused on two main categories: UI/UX design tools and AI-powered design tools. Both markets are experiencing rapid growth in the coming years.

At the same time, more product designers are turning to AI tools to streamline their workflow, moving beyond traditional software like Figma and Sketch.

Competitor Analysis

To identify gaps in the current landscape, we tested 10 leading AI tools across two domains: research and UI design.
Competitor's Weaknesses
Most tools only support research or design, not both. As a result, designs often miss key insights.
AI design tools usually offer only one solution at a time, with no alternative directions.
Editing AI-generated designs is time-consuming and inefficient.
Kami's Opportunities
All-in-one workflow, connecting research, design, prototyping, and testing.
Designs based on research, leading to more informed and relevant results.
Explore multiple options before choosing one, supporting creativity.
Editable AI outputs, easy to adjust and refine without starting from scratch.

Primary Research

To better understand designers’ current workflows, challenges, and attitudes toward AI tools, we conducted 10 in-depth interviews and collected 55 questionnaire responses.

Participants included designers from big tech companies, startups, and universities. .

70%

find it hard to synthesize large amounts of research data.

67%

find it difficult to make interfaces both beautiful and usable.

75%

think adjusting the results from AI design tools is difficult.

84%

think AI output quality is unreliable for design.

Design Goal

Based on the insights we gathered from designers, we identified clear needs across research, ideation, design, prototyping, and testing. Our goal was to streamline this entire workflow, making it faster, easier.
Gather and synthesize user insights
Generate high-quality, research-backed ideas
Design UIs aligned with user needs
Build prototypes with minimal effort
Iterate quickly with AI-powered feedback

Low-Fidelity Wireframes

The first version is divided into three sections: a layer panel on the left, the main design canvas in the center, and an AI panel on the right. In this version, we included only the essential features to ensure the interface is clean and easy to navigate.
Wireframe of the Design Canvas Page
Based on this design canvas page design, we built three user flows into a clickable wireframe prototype: initiating AI interaction, using AI for research, and generating designs with AI.

Usibility Test

We tested the three flows with over 30 designers from diverse backgrounds, gathered their feedback, and iterated on the design. This testing process went through more than three rounds.
Page for Newly Created Files
This is the screen shown after the user creates a new file. In the first version, the AI panel was on the right. Designers were confused and didn’t know what to do.

We changed it to a single input field in the center, making the prompt clear and reducing confusion.
Before
After
Page for AI-Generated Assets
The original design used a full-screen chat to show AI results, which felt unnecessary.

Now, all interactions happen in the AI panel, with updates shown directly on the canvas.
Before
After

Final Design

After multiple rounds of testing and iteration, we arrived at a streamlined layout that supports fast, flexible, and focused design work.
Assets & AI Panel

A transparent, chat-based AI partner that shows every step and supports fast, flexible iteration.

Properties Panel

All AI outputs remain fully customizable, giving designers complete control to tweak and perfect layouts effortlessly.

Prompt Bar

A simple entry point where designers can ask Kami to generate, upload, or style anything. Quick-start suggestions below help users begin right away.

Meaning of "Kami"

In Japanese, kami means both “paper” and “god”. Paper symbolizes creativity and new ideas; God reflects intelligent guidance. The name captures the blend of human imagination and smart AI support in design.

Reflection

Technical Details
To address the limited quality of UI outputs from existing AI design tools, we trained the model with a labeled dataset of diverse UI styles. Each design was tagged with its style name, which significantly improved the AI's ability to generate visually varied and higher-quality results.
Next Step
We’ll train the AI on a larger dataset of diverse UI styles. Then we’ll connect APIs from different tools to build a full product platform.
Main Takeaways
Testing really matters, people often behave differently than expected, and sometimes reveal valuable insights.

© 2025 Clayton Lin

Let’s collect the world’s together.

© 2025 Clayton Lin

Let’s collect the world’s
together.