An iOS app that helps non-Chinese-native parents and students study their books — turning printed pages into AI-powered interactive learning material they can understand, hear, and practice.
Pikabook takes book pages beyond the printed page. From a photo of a book page or a school handout, it generates line-by-line explanations, pinyin, audio clips, and flashcards for the words you don’t understand. And for the sentences you want to practice further, it offers pronunciation assessment.
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.
And it turned out it wasn’t only my problem. When I surveyed and interviewed other parents in the same situation, the picture was the same: the current experience is fragmented — a dictionary app here, a translation app there — and no single solution covers the whole 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.
Cursor · Firebase · Gemini · Google Cloud APIs · LLMs
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.
I kept an unfiltered build log on Threads ↗ while building and refactoring — the raw feelings behind this process. The morning Pikabook went live on the App Store, I woke up to the listing and posted in pure excitement. It was my first time shipping an app to a marketplace. It still felt unfinished — but it was real, and for a moment I just wanted to enjoy the small win with a cup of coffee.
Later I collaborated with a developer to improve the product further. With today’s AI models, almost everything feels possible at this product’s scale. Back then, though, I hit a wall that endless refactoring and reading the codebase with limited engineering knowledge couldn’t fix: subscriptions. My business model uses a usage-based trial after subscription — and I couldn’t tell where it was breaking. So I hired a developer to review the code from the ground up. That collaboration clarified what AI-assisted building still needs from experienced engineers: architecture, reliability, and production-level judgment — not screen-by-screen implementation.
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. That shift meant wearing the product manager / product owner hat far more than before. I spent more time in Cursor and Notion — roadmapping, writing the wiki, tracking bugs, documenting use cases for myself and the developer — than in Figma. Toward the end, I almost never opened Figma at all.
The harder lesson: customer reach matters more than I wanted to admit. A strong product means little if it never reaches the people who need it. Right after launch, fast marketing matters — and in Singapore, that was harder than expected. Sharing in Facebook groups and local social communities often got treated as advertising; after a few rejections and removals, I grew discouraged and under-promoted Pikabook. Losing that early momentum is still what I regret most.