Practet

Students preparing for English proficiency tests face a common gap. Traditional prep relies on classroom feedback that is slow and expensive, or generic practice papers that grade only the multiple-choice sections. The two highest-scoring components, Writing and Speaking, almost always end up self-reviewed or skipped entirely. Without honest, immediate feedback on those sections, and without a reliable estimate of their current band score, learners walk into the real exam under-prepared.

Practet closes that gap by putting an AI evaluator behind every Writing and Speaking attempt. Students get instant scoring, line-level corrections, pronunciation analysis, and band-score predictions that target 96% similarity to official results.

Supported tests

• IELTS (both Academic and General Training)
• PTE (both Academic and Core)
• TOEFL

What learners get

• Instant AI feedback on Writing and Speaking attempts
• Pronunciation analysis with phoneme-level scoring
• Adaptive study plans driven by weakness detection
• Time-limited authentic mock tests for every section
• 1M+ practice questions across Reading, Listening, Writing, and Speaking
• Band-score predictions before sitting the real exam

For institutions

• White-label portal with custom branding
• Bulk student management and group analytics
• Zero setup cost

How it was built

I designed and built the entire system end to end as the sole engineer. The frontend is a Next.js app written in TypeScript with Tailwind. The backend is a Django REST Framework API backed by PostgreSQL, with separate services for AI writing evaluation, speech-to-text, and pronunciation scoring. I built the adaptive study-plan engine, the test runner with strict timing and anti-cheat constraints, the analytics layer, and the multi-tenant model that powers the white-label offering.