thierryhenrry/kinyarwanda-ai-trainer

# Kinyarwanda AI Trainer Toolkit (KAT) A web app for Kinyarwanda language specialists doing AI training, annotation, and translation freelance work. Annotate faster, export in ML-ready formats, track job applications, and build a portfolio. ## Features - **Dashboard** — task stats, domain breakdown, quick actions, export - **Review workspace** — rate AI-generated Kinyarwanda (grammar, fluency, cultural accuracy), tag issues, correct text - **Translation editor** — EN ↔ RW post-editing with edit-distance tracking - **RLHF comparison** — pick Response A vs B with required rationale and rubric scores - **Job tracker** — DataLens Africa, Micro1, AuraOne, AfriCareers, LinkedIn, Digital Umuganda - **Portfolio** — summary of completed work with JSONL/CSV export - **Dark/light theme**, keyboard shortcut **Ctrl+S** to save, auto-persisted in browser localStorage ## Quick start ```bash npm install npm run dev ``` Open localhost. ## Deploy (Vercel) ```bash npm run build ``` Push to GitHub and import the repo on vercel.com. Share your deployed URL as a portfolio link when applying for AI trainer roles. ## Export format Tasks export as **JSONL** (standard for LLM fine-tuning) or **CSV**. Example review record: ```json { "id": "review_...", "type": "review", "source_lang": "rw", "original": "The patient should take medicine twice daily.", "ai_output": "...", "corrected": "...", "ratings": { "grammar": 4, "fluency": 5, "cultural_accuracy": 5 }, "tags": ["minor_grammar"] } ``` ## Earning with this tool 1. Complete sample tasks in each module to practice assessment-style work 2. Export JSONL samples for your application portfolio 3. Track applications on the Jobs page 4. Apply on platforms listed in the job tracker (remote Kinyarwanda AI trainer roles) ## Tech stack Next.js 16 · React 19 · TypeScript · Tailwind CSS 4 · localStorage