Voilo

Problem

Language learners need real-time spoken practice with feedback that adapts to their actual mistakes — not static, one-size-fits-all lessons.

Solution

Built the backend for a real-time AI language-learning app using FastAPI and WebSocket-based conversational pipelines powered by OpenAI's Realtime models. Added adaptive learning features — vocabulary tracking, mistake memory, and personalized difficulty adjustment — backed by PostgreSQL learning analytics, and deployed the scalable backend infrastructure behind a cross-platform mobile app.

Tech Stack

• FastAPI — Async-first backend to keep pace with real-time voice traffic.
• OpenAI Realtime API — Low-latency speech-to-speech conversation for live practice sessions.
• WebSockets — Persistent bidirectional connection for streaming conversational audio.
• PostgreSQL — Stores per-user vocabulary progress and mistake history for adaptive difficulty.

Challenges

Write-up coming soon.

Impact

Metrics coming soon.

Lessons Learned

Write-up coming soon.