All work
iOS · AI · Parent tool

Pikabook

AI-powered learning support, built from idea to release — helping non-Chinese-speaking parents turn a photo of school material into something their child can understand, hear, and practice.

Role
Founder, product, design & build
Platform
iOS
Year
2025 — 2026
Live
Pikabook app
Brief

Pikabook helps non-Chinese-speaking parents and students go beyond the printed page. With a photo of a worksheet, textbook page, or school handout, it generates pinyin, translation, AI voice playback, and pronunciation feedback — turning printed Chinese material into something parents and children can actually understand, hear, and practice.

The spark

From a parent’s frustration

Pikabook started from my own experience in Singapore. My son would bring home Chinese printouts, and I often could not understand them well enough to help. I could take a photo, ask GPT, or look things up manually — but the experience was fragmented. I needed pinyin, translation, pronunciation, audio, and a way to save useful content in one flow.

The bet: AI could help — but only if wrapped in a useful workflow. The value was not “OCR plus translation.” It was moving from “I don’t understand this page” to “I can help my child read, hear, save, and practice it.”

Not OCR plus translation — a complete learning workflow.

10+
Major revamps to ship v1
iOS
Released on App Store
1
Idea to release, AI-assisted
Stack

Cursor · Firebase · Gemini · Google Cloud APIs · LLMs

From idea to release

AI-assisted product building

I designed the product, defined requirements, prioritized the roadmap, and used AI coding tools to build and release the first version — before today’s coding agents became mainstream. That required constant product judgment: breaking features down, checking implementation quality, enforcing structure, and repeatedly cleaning up architecture around state management, caching, and feature boundaries.

Later I collaborated with a developer to improve the product further. That raised an important question: when a designer can prototype, specify, build, and ship with AI, what kind of developer collaboration is still needed? Not “no developers” — but a shift. Developers become most valuable for architecture, reliability, scalability, security, and production-level judgment — not screen-by-screen implementation.

What I built
  1. 01Photo-based Chinese text extraction with pinyin, translation, and AI voice playback
  2. 02Pronunciation feedback, saved notes, and early flashcard / review flows
  3. 03Onboarding, usage limits, and subscription thinking
  4. 04Product direction, UX design, roadmap, release decisions, and developer handoff
Key decisions
  1. 01Build around the real parent moment. A parent sitting with a child and a printed Chinese page — the flow had to be practical, fast, and forgiving, not a full curriculum platform.
  2. 02Go beyond translation. A translated answer is useful once. I designed for pinyin, voice, pronunciation, saving, and review so the printed page could become reusable practice material.
  3. 03Use AI to ship, but not blindly. AI accelerated building, but also created technical debt. I repeatedly enforced clearer architecture and feature boundaries to keep the app maintainable.
What I learned

A powerful AI feature is not the same as a sharp product. OCR, translation, pinyin, voice, and pronunciation feedback only matter when they support a real user routine — which is what led me toward No Worry Tingxie’s narrower weekly use case.

AI can dramatically shorten the distance between idea and release, especially when the product is defined clearly. But speed creates its own mess — without strong product judgment and architectural discipline, AI-generated code becomes hard to maintain.

Pikabook is where I moved from product designer to product builder: finding the use case, defining the workflow, using AI to ship quickly, collaborating with developers where it matters, and deciding what the first version should and should not be.

Back to work
All projects