Embeddings

Beginner — Explain an embedding.

• It is a numerical vector that represents useful features of content, often meaning.
• Similar wording or meaning tends to produce nearby vectors, enabling retrieval of “vehicle collision” for a query about “car accident.”

Intermediate — What does cosine similarity tell you?

• It compares vector direction and helps rank related content.
• A high score is not a probability that the passage is true or sufficient.
• I use a compatible metric and calibrate thresholds on domain-specific queries.

Senior — How do you upgrade embedding models?

• I version models and index data, re-embed the intended corpus into a separate compatible space, evaluate filtered retrieval, and switch query/index routing together.
• I preserve source data and support rollback instead of mixing arbitrary old and new vectors.

Scenario — Policy-code search fails despite good semantic search. What do you do?

• I add exact identifier fields and lexical/hybrid retrieval.
• Embeddings may not preserve rare codes or precise numbers well.
• I test actual identifiers and keep them intact during query rewriting and text preprocessing.