How YouTube interviews

1. The shape of a YouTube loop

YouTube hires under Google's process. The committee, the levels, the rubric — all Google. But the questions feel different. YouTube engineers spend their days on media pipelines, recommendation infrastructure, and planet-scale CDN problems. The system design rounds usually go there.

A typical senior IC loop:

Round 1   Coding / problem solving           (45 min)
Round 2   System design (media-pipeline)     (45 min)
Round 3   System design (general distributed) (45 min)
Round 4   General cognitive / Googleyness    (45 min)
Round 5   Cross-functional (bar-equivalent)  (45 min)

The hiring committee makes the call after the loop. Same as Google. The bar is set by Google-wide criteria; YouTube's hiring team can advocate but doesn't decide.

2. What YouTube grades you on

Same four Google signals — General Cognitive Ability, Role-Related Knowledge, Leadership, Googleyness — applied through the media lens.

In a YouTube system design round, the role-related knowledge looks like:

YouTube interviewers expect the candidate to reach for media-specific primitives by name. "Adaptive bitrate streaming over HLS" lands better than "chunked video."

3. What's different here

A few patterns specific to YouTube:

• The scale is genuinely planetary. Most "design system X" conversations end at "millions of users." YouTube's scale is billions of hours watched per day. The interviewer expects you to scale your numbers up accordingly.
• Egress is the cost driver. Computer scientists love compute; YouTube is dominated by network egress. Designs that ignore egress cost lose at the L6 bar.
• Media isn't just files. A "video" is many things — many bitrate variants, many languages of subtitles, captions, thumbnails, posters, audio tracks. Designs that treat a video as one blob miss the complexity.
• Cold-start matters for recommendations. A new channel has no view history. The recommendation system has to surface their content somehow. Interviewers ask about this directly.
• Comments are at video-content scale. Hundreds of thousands of comments on a viral video. The substrate is heavier than most chat systems.

4. The level bar

Same Google ladder:

• L3 / L4 (entry / junior senior). One system design. Scope and competent components. Single deep dive sufficient.
• L5 (senior). Two system design rounds. One deep dive at level on at least one of them.
• L6 (staff). Two deep dives per round. Anti-patterns named. Multi-region / multi-CDN reasoning.
• L7+ (senior staff / principal). Roadmap conversations. Cross-system reasoning. Migration plans on top of YouTube's existing architecture.

This chapter targets L5 with explicit L6 notes where the bar shifts.

5. The YouTube-flavored mistakes

6. The seven scenarios + take-home in this chapter

7. The next page

The video pipeline. From the moment the user clicks upload to the moment the video is playable on a phone in another country. We start with the upload itself — chunked, resumable, multi-region — then walk through encoding and storage.