Interview Questions › AI Product

AI maths tutor for school students

Design an AI tutor that helps class 8 to 10 students with maths homework without just giving them the answers.

Context

Parents pay. Teachers worry about cheating. Models sometimes make arithmetic mistakes.

Clarifying questions to ask first

• Who is the buyer and who is the user?
• How do we measure learning?
• What happens when the model is wrong?

How to approach it

1. Start from the user problem, not the model: is AI the right tool here?
2. Describe the experience and where the model fits in it
3. Define what "good output" means and how you would evaluate it before and after launch
4. Name the failure modes (wrong answers, bias, privacy, cost) and how you contain them
5. Close with a launch plan and success metrics

Framework: User and problem -> Is AI the right tool? -> Experience design and human-in-the-loop -> Quality bar and evaluation -> Risks, cost and latency -> Staged launch and metrics

Sample answer

• Socratic hints, step checking rather than final answers
• Verify maths with a calculator or solver, not the model alone
• Parent view of progress

1. Accuracy eval on a bank of solved problems
2. Pre/post quiz scores versus a control group
3. Monitor answer-only requests and cheating patterns

Hint-first tutoring with a maths engine checking every step. Success is improvement on independent quizzes compared with a control group; maths accuracy is a launch gate.

What interviewers look for

• Learning gains
• Session completion
• Maths accuracy