Projects / AI products
AI language tutor.
An AI tutor built into an English-language curriculum and its mobile app. It answers questions inside the lesson the learner is on, corrects and explains at their level, and gives them a safe place to practise conversation.
- Stack
- Python, Django, LLM APIs, React Native, speech assessment service
- Scale
- Every lesson in the curriculum, web and mobile
- Role
- Tech lead, built the service
Flow.
- Learner asks
- Lesson context
- Level & history
- Guardrails
- LLM
- Explanation or correction
- Progress
How it works.
The tutor lives inside the lesson, not in a separate chat window. Every request carries the lesson the learner is on, their level, and what they have recently struggled with, so an answer about the present simple is about the present simple in this lesson, at this level.
Written answers come back corrected with a short explanation of the rule. Pronunciation practice sends the recording to the self-hosted speech assessment service and returns a score with what to fix. Conversation practice keeps the learner talking and corrects lightly as they go.
The service is a Python and Django backend that the web platform and the React Native app both call, with guardrails that keep the tutor on the curriculum and on task.
Hard parts.
- Keeping answers grounded to the curriculum and pitched at the learner’s level instead of whatever the model would say by default.
- Latency inside a lesson: a tutor that takes ten seconds breaks the flow of learning.
- Learners who ask about English in their own language, and need an answer that teaches instead of one that translates.
Outcome.
Shipped as the AI Tutor tab in the mobile app and across the web courses, giving every learner on-demand help without waiting for a teacher.
Screens.