Design AI features for YouTube (discovery, creators, and moderation).

### Signal to interviewer

I can design platform AI using multi-sided incentives instead of single-metric optimization.

### Clarify

I would clarify product priorities across viewer retention, creator growth, and moderation risk tolerance.

### Approach

Apply a balance matrix: discovery ranking, creator copilot features, and safety moderation integrated through shared policy and feedback loops.

### Metrics & instrumentation

Primary metric: quality-adjusted watch-time. Secondary metrics: creator growth consistency, recommendation diversity, and moderation precision/recall. Guardrails: harmful exposure rate, creator fairness imbalance, and false takedown volume.

### Tradeoffs

Aggressive personalization boosts engagement but narrows exploration. Strict moderation lowers risk but can suppress legitimate content.

### Risks & mitigations

Risk: recommendation echo chambers; mitigate with diversity quotas. Risk: creator distrust in moderation; mitigate with explainable decisions. Risk: abusive content evasion; mitigate with adaptive classifiers plus human review.

### Example

For educational channels, AI boosts discoverability through topic-intent matching while moderation distinguishes expert critique from harmful misinformation patterns.

### 90-second version

Design YouTube AI as a balanced ecosystem engine: optimize discovery, empower creators, and enforce safety with transparent guardrails and multi-objective metrics.