Surface-level understanding
Students received the correct answer quickly but often missed the logic or evidence behind it.
Build trust and transparency
Show verified academic sources so students can understand where the AI explanation comes from.
Gauth by ByteDance · Shipped 2024
Helping students learn, not just get answers
In 2024, I worked as a UX researcher and product designer at ByteDance’s AI division, Gauth, translating student research into two shipped product directions.
- Role
- UX Researcher
Product Designer - Timeline
- Sep–Dec
2024 - Scope
- Research · Strategy
UX/UI · PRD - Team
- Product · Design
Engineering
An AI learning app used by 4 million students worldwide.
Gauth helps middle-school, high-school, and college students solve academic problems through instant feedback, step-by-step explanations, and personalized recommendations. The goal was to shorten problem-solving time while supporting independent learning and critical thinking.
Reach
4M
Research
50+
Post-launch
+10%
My contribution
I led discovery through product definition, turning 50+ student interviews into two shipped product directions.
Understand where fast answers stopped supporting learning.
The research focused on the moments when students moved from solving a problem quickly to actually understanding it.
Learner context
Workload, confidence, and the pressure behind instant help.
Learning moment
Where students reviewed reasoning or stopped engaging.
Trust signal
What made AI guidance understandable and verifiable.
Observed journey
Capture
Review
Follow up
One learning loop revealed where speed helped and where understanding broke down.
Students wanted quick help, but trusted explanations they could follow.
Across 50+ interviews, the same tension appeared: efficiency brought students in, but transparency and structure determined whether they kept learning.
Students did not need more answers. They needed evidence they could trust and practice they could return to.
Two learning breakdowns shaped the product strategy.
The opportunity was not simply to provide more information, but to help students understand where an answer came from and how their learning was progressing.
Disorganized learning
Scattered practice made it difficult to focus on weak areas or understand progress by topic.
Structure by concept
Organize ACT and SAT questions around key concepts to support systematic practice and mastery tracking.
Solution
01
Smart Source Retrieval
Make AI reasoning easier to verify.
Smart Source Retrieval added multiple layers of verification for humanities subjects. Gauth retrieves academic sources, displays supporting evidence beside the answer, and connects reasoning with historical or conceptual context.
Verified sources
Students can open references and understand where an explanation comes from.
Context, not just answers
Guided breakdowns connect facts with the historical or conceptual background around them.
Solution
02
Concept-based Test Prep
Turn test practice into a learning loop.
Concept-based Test Prep organizes ACT and SAT content around key topics, then provides reflective, step-by-step feedback after each mistake to strengthen comprehension and long-term confidence.
Explain the mistake
Clarifying why an answer is correct builds understanding instead of encouraging memorization.
Show progress
Visible progress and supportive feedback create momentum when students feel stuck.
Research moved into the product, not just the presentation.
The work was presented across the team, translated into product requirements, and shipped as part of the next iteration.
Adoption
4 million global users
Engagement
10% DAU increase after the upgrade
Market
#1 Education app in the US App Store
08 · Learn · Takeaways
01
02
03
Responsible AI design begins with how people learn.
Understand EdTech behavior
I learned how students interact with educational tools and how to design for real learning contexts.
Balance utility and ethics
AI education products should support reflection and confidence rather than simply making answers faster.
Translate research together
Close collaboration with design, engineering, and product turned qualitative insight into aligned product decisions.