Design an AI assistant for students to study and practice.

### Signal to interviewer

I can design student AI for learning efficacy and retention, not shortcut-driven completion.

### Clarify

I would clarify grade level, subject scope, exam goals, and teacher/parent involvement expectations.

### Approach

Implement a practice-adaptive tutor loop: diagnose, assign targeted practice, deliver tiered hints, and schedule spaced reinforcement.

### Metrics & instrumentation

Primary metric: session-over-session accuracy improvement on target skills. Secondary metrics: retention checkpoint pass rates, completion consistency, and confidence calibration. Guardrails: hint dependency, frustration drop-off, and shallow answer copying.

### Tradeoffs

Richer tutoring support accelerates progress but may reduce independent problem-solving if overused. Lower guidance preserves struggle but can increase drop-off.

### Risks & mitigations

Risk: answer farming behavior; mitigate with attempt-gated hints. Risk: misleveled content; mitigate with adaptive difficulty calibration. Risk: low transfer to exams; mitigate with mixed-context practice.

### Example

For chemistry prep, AI detects stoichiometry errors, delivers focused drills, and later tests transfer through multi-step reaction problems.

### 90-second version

Design the assistant as a mastery coach with adaptive practice and fading support. Optimize long-term learning gains while controlling hint dependency and dropout.