Problem
Students and researchers need academic text to sound natural without losing technical meaning, discipline-specific terminology, or personal writing voice.
Back to projects
AI Product Engineering
Polishly Academic Writing Refinement SaaS
Bilingual academic writing refinement platform for Chinese and English drafts, designed to reduce AI-sounding phrasing while preserving meaning, terminology, and author voice.
AI Product Engineering
EN + 中文
Languages
8
Disciplines
3 levels
Intensity
3/day
Quota
Evidence mode
Metrics snapshot
EN + 中文
Languages
8
Disciplines
3 levels
Intensity
3/day
Quota
Next.js 16React 19Gemini FlashDiff UXDOCX Export
Architecture
- 01Editor-first homepage with split-panel review workflow
- 02Gemini Flash REST route handler with discipline-aware CN/EN prompts
- 03Structured response parsing into editable JSON change sets
- 04Character and word-level diff rendering with accept/reject controls
- 05DOCX export with colored insertion/deletion runs and lightweight i18n
Case study
Contribution
Designed the product concept and implemented the core editor, bilingual prompt system, diff workflow, term locking, quota behavior, and export path.
Engineering
Used Next.js 16 App Router, React 19, TypeScript, Tailwind CSS 4, the diff package, docx generation, cookie-based free quota, and Gemini Flash API integration without an SDK.
Outcome
Created a reviewable MVP that demonstrates AI product thinking, LLM integration, document tooling, bilingual UX, and product roadmap planning for auth, payments, durable rate limiting, analytics, and extensions.