Brief

Product Goal

User Research

Competitive Analysis

Final Outcome

Frame Reuse

Work

About

Resume

Bytedance Intern

My internship experience at Bytedance to help build the AI-powered Q&A Asssistant.

As the Douyin platform explores how AI can enhance daily work efficiency, I was tasked with developing the Douyin R&D Assistant to improve the early PRD analysis process. Through this project, I have explored the functionality and framework for the AI-powered Q&A products, which has also been applied to other similar internal products I have taken charge of.
Internship
2025/1
3 months
UIUX Design
Information Architecture
Researching
Visual Design

Product Goal

The less connection among remote workers

Research have shown that around 53% workers feel a lack of connection and belonging with other colleagues. Initially, it might feel great to have so much flexibility and independence in a remote position, but over time, that level of isolation often leads to remote work loneliness.

Develop an efficiency-focused LLM Q&A platform for the Research and Develop department.

Challenges while treating customers with food restrictions

What kind of platforms they would like to collaborate with

Streamline PRD requirement integration
Improve the efficiency of developing process
Facilitate knowledge sharing for Douyin R&D

User Research

Developers

Age:   20–40 years old, predominantly male.

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

Painpoint #1

Time-consuming process of the early-stage PRD analysis

 The creation and planning of PRD in the early stages are time-consuming and labor-intensive, with limited focus. This results in incomplete requirements can weaken the final output of the developing process.

Painpoint #2

Low reuse rate of existing knowledge assets and historical incidents

 Existing historical problems are scattered across different documents and records, making them hard to archive effectively. Developers lack quick access to solutions when facing similar problems.

Maybe... AI could possibly help developers in the early PRD analysis stage.

Problem statement

How might we relieve the isolation among remote workers?

The increasing number of remote workers

It is expected that they will be 87% increase of remote work from pre-pandemic levels to 2025

The increasing number of remote workers

It is expected that they will be 87% increase of remote work from pre-pandemic levels to 2025

Our Mission

How Might we build a platform that aligns with the R&D workflow and enables efficient reuse of historical solutions; applicable for use in the early PRD refinement stage to reduce costs and increase efficiency.

Competitive Analysis

In order to better understand the existing LLM products, I have conducted a competitive analysis focusing on their functionality and framework.

General Apowered Q engine

  • Designed for users with needs for root cause analysis and knowledge Q&A. It generates query results by gathering information from diverse related sources using GPT and in-house model
  • Provides intuitive interaction and visually appealing design, earning strong user feedback and reputation.

Internally compatible Q&A engine

  • Features robust internal data integration, seamlessly incorporating Lark documents, chat records, knowledge bases, and other internal content.
  • It supports the selection of different information sources based on the initial content, enabling further improvement in accuracy and relevance.

Light

&

Remote Workers

Light is an essential tool for remote workers, and it serves as a common representative for 'living places' and 'home'. Besides, it is an element that has large potential to make the visual identity unique and appealing. For these three reasons, we chose light as our key concept element.

Interviews

The demand for flexible working hours.

Remote work offers greater flexibility in terms of when and where work is performed. It greatly improves work-life balance.

A need of belonging

Remote workers may feel disconnected from the company's culture, values, and mission as they are not physically present to experience or participate in cultural events and activities.

The feeling of isolation and loneliness.

Remote workers are not physically present with their colleagues, which can lead to feelings of isolation and disconnection.

Based on the research findings, I have developed insights into a potential framework that our product should adopt to address specific and professional demands.

Final Outcome

New Threads

The home page of the product, providing users with an easy entrance to raise questions. Related questions are provided below to inspire ideas.

Result Page

The answer structure includes: Searching Process + Sources + Answers. Among the paragraph, there also displays the cited sources.

Highlight #1

Document Links Tags

The input frame allows the pasting of Lark document to analyze internal PRD. In order to help user better edit the link, I have contained it in a tag and added a close button.

Highlight #2

Diverse Types of Data

In order to provide a comprehensive analysis to the PRD, I have designed 3 different types of data to visualize the possible information developers may need.

  • Chart: Collapse by default to control the length of responses. Provide expanding button for users to view more details by their choice.
  • Text: In Markdown format.
  • Coding blocks: Follow the color coding and format of other development softwares.

Highlight #3

Anchor Bar

Since the answer of our product may be in a long paragraph for its large amount of data, I have designed an anchor bar to highlight each part of the answer, to help users better find the answers they need.

  • Default: Each dot labels a single H1 in the paragraph, enabling users to roughly understand the overall length of the answer.
  • Hovering: The bar will expand to showcase a more comprehensive information, allowing users to quickly jump to the contents they are interested in.

Framework Reuse

After becoming familiar with the information architecture of AI search, I was able to quickly design similar AI-powered Q&A products, improving my work efficiency. Additionally, my research on competitors and analysis of their frameworks helped the team accumulate relevant product insights.

Here is a product, MagicSurvey, built by reusing the framework and insights.