vector DB

A numerical vector that represents the meaning of a piece of text (or image, audio, etc.). Two embeddings that are "close" have similar meaning. The substrate of semantic search and RAG. Stored in pgvector, Pinecone, or similar.

When your vector embedding setup produces results that are obviously, embarrassingly wrong — similar items that should cluster together are nowhere near each other, or wildly unrelated things are returning as top matches. An embedding fail is the moment you demo your semantic search and it surfaces something deeply unrelated as the top result. These moments are both humbling and informative, usually exposing that your embedding model wasn't the right fit for your domain or your training data was messier than assumed.

Someone with an almost supernatural command of vector embeddings — the person on the ML team who intuitively understands high-dimensional semantic space and consistently produces embedding pipelines that just work beautifully. An embedding god can look at a similarity matrix and immediately diagnose what the model is misunderstanding. They're the rare engineer who has internalized the math well enough to reason about it like intuition, and the team always tags them when the retrieval system starts acting weird.

Tying an AI's output to verified external facts (documents, databases, search results) to reduce hallucination. RAG is the most common technique.

In machine learning: a numerical representation of text/images/etc. as a vector in high-dimensional space. The technical building block of modern LLM systems.