← All case studies

A real-time AI voice tutor that adapts to how each child learns.

Adaptive voice tutoring for children aged 5 to 11, with animated tutors that listen, respond and build a learning profile for each student.

Voice latency target for live conversation (reported)< 500ms
Learner archetypes used to personalise teaching10

Built as a senior engineer at Algotix AI · Visit teachers-pet.net ↗

Young children lose interest fast. The tutor had to respond in near real time and adjust to each child, not read from a script.

  1. Real-time voice streaming with the OpenAI Realtime API and four ElevenLabs voice agents behind an animated Rive avatar.
  2. Student knowledge profiles that place each child in one of 10 learner archetypes, refined with GPT post-processing.
  3. Redis caching and cross-device sync so a session can continue on another device.
  4. Resumable S3 uploads and an asynchronous transcription pipeline that keeps the live session fast.
Next.jsTypeScriptOpenAI Realtime APIElevenLabsDynamoDBAWS / SST

The matching package.

Have a problem like this?

Tell me what is slowing you down. I will reply with the clearest next step.