PrepBox - Adaptive Math Tutoring Platform

Lead Product Designer & UX Researcher

Duration: Feb 01, 2021 - May 31, 2025 ( 3.4 Years)
Client: MGL Learning (PrepAnywhere Inc.)
Platform: Mobile & iPad (iOS, Android)

Team: Jina Seok, Mike Woo, Nesan Vettivel, Roger Takeshita, Amy Song, Pankaj Punia

Context & Problem Framing

Mathematics education in the K-12 sector has long suffered from a structural mismatch: curriculum delivery is predominantly linear and one-size-fits-all, yet student learning is non-linear, highly individual, and deeply affected by prior knowledge gaps. MGL Learning engaged me to help build a platform that would not just digitise existing instruction, but fundamentally reimagine how personalised learning could work at scale.

The initial research challenge was to understand what 'adaptive learning' meant in practice for students, teachers, and tutors - and to identify where existing digital tools were failing them.

Research Questions

Where do K-12 students most commonly lose comprehension in mathematics, and what do they do when they reach those points?
How do teachers and tutors currently identify student knowledge gaps, and what information do they wish they had?
What affordances of mobile and tablet devices are most compatible with active problem-solving (as opposed to passive consumption)?
What friction exists in current tutoring workflows, and how could a platform reduce it while maintaining human connection?

Goals

Create personalized learning paths for K-12 students that adapt to their strengths, gaps, and preferred features (AI + human tutor support).
Enable active learning: students should not just consume lectures but do work - write steps, solve problems, see where they error.
Deliver strong educational scaffolding: lectures, hints, solution videos, instant feedback, 1-on-1 or on-demand tutoring when needed.
Make the platform engaging and usable on iPads (widely used in target schools) - ensuring UI/UX is optimized for pen input, whiteboard-style work, etc.
Drive adoption, retention, and revenue — especially via premium AI plan, reduce friction in support, and scale to new users/countries.

My Approach & Design Process

User Research & Assessment

Engaged with students, teachers, and tutors to understand pain points: where students get stuck, what feedback they need, how much human help vs. system help is useful.
Analyzed existing platforms and curriculum alignment (grade level, standards) to identify gaps and requirements.

Information Architecture & Learning Paths

Designed a flow where students begin with an assessment to benchmark their math skills.
Based on assessment, the system assigns learning resources (lectures, problem sets) tailored to their grade and skill gaps.
Enabled switching between assessment and learning so students can revisit concepts as needed.

Interaction Design: Doing, Not Just Watching

Developed a whiteboard component: students write out full steps, show their reasoning, not just pick answers.
Incorporated tools like pen input on iPad, so the solving feels natural and hands-on.
Designed UI to support hints, scaffolded support, and feedback - lecture videos, solution walkthroughs, instant AI grading.

Scaffolding & Support Features

Layered support: small lectures, hints, full solution videos, and on-demand human tutor help.
Real-time feedback: instant scoring/grading from AI to help students self-correct.
Content monitoring: ensuring that video content, problem difficulty, and system suggestions map appropriately to student performance.

Iteration, Usability & Visual Design

Created wireframes, prototypes, and tested them with students to ensure comprehension and ease of use.
Optimized for iPad usage: touch, pen, whiteboard interactions, screen flow.
Established visual consistency (typography, icons, spacing) and intuitive UI for navigation between content, assessment, feedback, and tutor support.

Key Design Decisions

Whiteboard-style problem solving vs. multiple-choice / button-click only approach. This was to encourage showing work and deeper understanding.
Assessment first + adaptive path, rather than linear textbook progression, so each student gets what they need—not what the curriculum assumes.
Async + human tutor backup, so even when students are on their own, they’re not stuck indefinitely.
Instant feedback via AI, to tackle repeated mistakes quickly and reinforce correct reasoning.

Impact & Results

Internal data shows 65% increase in revenue following the implementation of the mobile platform and its AI plan.
Strong growth in user base: 600,000+ new users after U.S. expansion - total reach grew to ≈58.25 million users.
8× increase in AI plan sales - indicating that users found value in the premium features (feedback / tutoring).
Raised $5M+ in funding, supported by improved product offerings & clearer product-market fit.

Metrics, revenue, users, funding info by CEO Min Lee

Reflection & Lessons Learned

Personalization + feedback loops matter: letting students see where they failed, and giving them ways to correct in context, improves engagement and understanding.
Tool fidelity is important: when you allow natural input (whiteboard, pen, writing out work), you align closer with how students think and learn.
Balancing scale with quality: we needed to produce a huge amount of content, but always ensure it was scaffolded, clear, and useful—not just “more stuff.”
Team coordination & continuous iteration were critical — content, AI, UX, visual, and tutoring features all needed to evolve together as users used the product and gave feedback.