Embedding
An embedding is a list of numbers that captures the meaning of a piece of text, image or audio, so that content with similar meaning gets similar numbers and the computer can measure closeness with plain arithmetic.
No jargon, no math
In plain words, what is it?
An embedding model reads a piece of content and compresses its meaning into a few hundred numbers, a vector. Think of it as a postal address for meaning: “hydraulic pump seal” and “oil leak gasket kit” land in the same neighbourhood even though they share no words. The model is trained so that distance in this number space means difference in meaning: cosine similarity (the standard measure of the angle between two vectors) scores how close two pieces are. Embeddings are the raw material of semantic search and vector databases: once every document has an address, finding related documents becomes a geometry problem instead of a word-matching problem. The catch: the addresses only make sense inside the model that created them, and they carry no facts on their own, only resemblance.
- 01
Content in. Text, an image or audio enters the embedding model.
- 02
Compression. The model distils its meaning into a few hundred numbers.
- 03
The vector. The numbers form the embedding: the content's address in meaning space.
- 04
Similarity. Close numbers mean close meaning, measured by cosine similarity.
- 05
Storage. The vector lands in a vector database, ready for similarity search.
Terms that live nearby.
Vector Database
A vector database is a database built to store numeric fingerprints of content (embeddings) and find the closest ones to any query in milliseconds, powering semantic search, recommendations and AI retrieval.
ReadSemantic Search
Semantic search finds content by meaning instead of exact words: the query and the documents are converted into embeddings, and the closest matches are returned even when no keyword overlaps.
ReadChunking
Chunking is the way long documents are cut into pieces before indexing: the cuts decide what the search can find, so a good chunk is a self-contained passage that keeps its meaning when read alone.
Read§ 04
Want this running in your stack?
We design and ship AI systems with permissions, logs and human approval. It starts with a 30-minute scoping call.