Skip to content
Coding FodingcodingFoding(you: Human)
  • Careers
  • Pricing

Theme

Heading font

  1. Home
  2. AI & LLM Engineering
  3. LLM Internals
  4. Self-Attention

Part 04 · LLM Internals

Self-Attention

Query, Key, ValueScaled dot-productCausal maskMulti-head attention

The heart of the transformer, worked out with real numbers: queries, keys and values, the causal mask that stops a word peeking ahead, and why many heads beat one.

⏱ 2-3 hours · Station 26 of 92

252627
← Previous stopEmbeddings and Positional EncodingNext stop →Transformer Architecture
Coding FodingcodingFoding(you: Human)

Everything you need, in one path.

Start free

Learn

  • Backend First Principles
  • DevOps & Cloud
  • System Design
  • AI & LLM Engineering
  • Database Internals
  • Frontend System DesignSoon

Practice

  • DSA problems
  • AI / ML problemsSoon
  • System Design Problems
  • Build to Hire

Interview

  • Interview PrepSoon
  • Most Common QuestionsSoon
  • Interview ExperiencesSoon
  • Success StoriesSoon

Community

  • Leaderboard
  • Blogs
  • Roadmap

More

  • Careers
  • Pricing

//codingFoding()

© 2026 Coding Foding · Made with ❤️ in Bangladesh