TikTok RD Assistant

An efficiency-focused AI platform for TikTok’s development team, to streamline knowledge access and daily workflows.

My Role

My Role

UX Design Intern

Responsibility

Responsibility

AI Design, Visual Design, User Research

Team

Team

UX Designers,

Product Managers,

Software Engineers

Duration

Duration

2024/10-2025/1

3 months

The Mission

Developed an efficiency-focused LLM Q&A platform that accelerates early PRD refinement by reusing workflow-aligned knowledge for internal developers.

The Impact

Used by 4+ developer teams across TitTok.

LLM framework scaled to support 2 internal AI product lines.

Framework integration delivered 15% MAU growth in a second AI product.

Used by 4+ teams across TitTok.

LLM framework scaled to support 2 internal AI product lines.

Framework integration drove 15% MAU growth in a second AI product.

Used by 4+ developer teams across TitTok.

LLM framework scaled to support 2 internal AI product lines.

Framework integration drove 15% MAU growth in the second AI product.

Used by 4+ developer teams across TitTok.

LLM framework scaled to support 2 AI product lines.

Framework integration delivered 15% MAU growth in a second AI product.

Used by 4+ teams across TitTok.

LLM framework scaled to support 2 internal AI product lines.

Framework integration drove 15% MAU growth in a second AI product.

The Outcome

The Outcome

The Outcome

The Outcome

The Outcome

Home Page

An inviting entry point for curiosity and idea discovery


An inviting entry point for curiosity and idea discovery

Answer Page

A structured, source-backed answer flow with quick navigation.

The Process

The Process

The Process

The Process

User Research

To uncover key opportunities, I investigated developers’ workflows to identify pain points in early PRD analysis and knowledge reuse.

User Research

To uncover key opportunities, I investigated developers’ workflows to identify pain points in early PRD analysis and knowledge reuse.

Developers

Age: 20–40 years old, predominantly male.

Traits: Young, tech-driven, working in a fast-paced environment.

Painpoint 01

Time-intensive early PRD analysis

Time-consuming and unfocused early PRD planning often results in incomplete requirements and weaker development results.

Painpoint 02

Limited knowledge and insights reuse

Historical issues are scattered across documents, making archiving difficult and limiting developers’ access to past solutions.

User Senario Map

Then I created an ideal workflow showing how AI supports developers’ tasks.

User Senario Map

Then I created an ideal workflow showing how AI supports developers’ tasks.

The insights - IA

Based on the competitive review, I defined the information hierarchy for our Specific and Professional Q&A platform

The insights - IA

Based on the competitive review, I defined the information hierarchy for our Specific and Professional Q&A platform

TikTok RD Assistant

Specific and Professional

-Composite answers for complex analyses.

-Comprehensive search for advanced validation.

-Delivers basic, straightforward Q&A.

Perplexity

Extensive and General

-For text-based, daily academic searches.

-A simple flow for casual and professional use.

-Adds broader, advanced features.

However, the composite answer has brought a problem

Challenge - Usability Testing

How to balance the depth of composite information with ease of navigation.

Challenge - Usability Testing

How to balance the depth of composite information with ease of navigation.

Searching sources

Before

Overly text-heavy, leading to inefficient source communication.

After

Achieve a cleaner layout by displaying only the most essential sources.

Anchor Bar

Before

Requires users to scroll through the long paragraph of information.

After

Provides an anchor bar that highlights key sections for quick navigation.

Chart Design

Before

Overuse of abbreviations may hide key information.

After

Offers flexible line abbreviation for a balance of clarity and simplicity.

Impact

Impact

Impact

Impact

4+

4+

developer teams adopted the framework across TikTok

2

2

internal AI product lines scaled on the LLM framework.

15%

15%

MAU increase in the second AI product integrated this framwork.

MagicSurvey — Built on the LLM Framework

Takeways

Takeways

Takeways

Takeways

Being a Proactive Learner

It's common to design for a product you've never used, or a field you're not familiar with. Always stay curious and proactive in learning to prepare for new knowledge.

Listen Carefully, Speak Boldly

Listen Carefully,
Speak Boldly

Listen Carefully,
Speak Boldly

It's common to design for a product you've never used, or a field you're not familiar with. Always stay curious and proactive in learning to prepare for new knowledge.

↓ Photos with my amazing team — grateful for them all.