How TikTok Learns Your Personality in 30 Minutes (Machine Learning Explained)

TikTok doesn't read your mind. It reads your behaviour — and turns it into mathematics. In this video we break down the full machine learning stack behind the For You Page: from raw watch-time signals, to high-dimensional embeddings, to collaborative filtering, deep neural ranking, and a real-time feedback loop that updates in under a second.

This is how a billion users get served personalised feeds with no human curation involved.

In this video:
- The exact signals TikTok measures (and why negative signals matter as much as likes)
- What embeddings are and how your watch history becomes a point in 10,000-dimensional space
- How collaborative filtering finds "users like you" without ever asking who you are
- The two-stage retrieval + ranking funnel that runs in under 200ms
- Explore vs exploit: why TikTok deliberately shows you things you didn't ask for
- How the model retrains on your actions in under 1 second

No code. No jargon. Just clear diagrams.

Subscribe to SysSketch — programming systems, visualized.

Видео How TikTok Learns Your Personality in 30 Minutes (Machine Learning Explained) канала SysSketch