Helping people preparing for IELTS practise speaking. The idea came from a friend; my job was to turn it into something usable and useful from a designer’s point of view, and I wrote the front end.
Who it’s for
This was someone else’s idea, and now the job is to make it usable, and useful, from a designer’s perspective.
The users are Chinese test-takers preparing for IELTS. When answering a speaking question, many people first think in Chinese and then translate, so they get stuck on words they can’t say:
Describe the clothes you like to wear: jeans, hoodies, jackets. These words rarely appear in vocabulary lists, but you really need them when speaking.
The idea
- Help people turn what they mean in Chinese into natural English. Collected over time, these phrasings become a personal phrase bank, generated by AI;
- find the words a person keeps using when they speak, and focus on those.
What we built
Work started on May 2, 2025, and there was a working version the same day:
- Speaking practice: IELTS question types, including Part 2 topic cards and a focused “describe a place” set;
- Recording and playback: listen back to your answer;
- Analysis and rewrite: the AI acts like an IELTS examiner. It first fills in any parts of the question you missed, then polishes fluency, vocabulary, grammar and structure while keeping your own words where it can. It ends with a table: what you said, the upgraded version, and why;
- Vocabulary: save the phrasings worth learning, backed by a small separate API.
Tech
Next.js and Supabase for sign-in and data, with the prompts kept in the project. The vocabulary API runs on Azure Functions. The first UI was generated with v0 and then reworked by hand.