Foundations · Established · Intermediate

Embeddings

Also known as: Vector embeddings

Numeric vector representations of text, images or other data where distance corresponds to semantic similarity.

What Embeddings is

An embedding turns meaning into geometry. Two paragraphs about refund policy land near each other even with no shared words, which is what makes semantic search possible.

How it works

An encoder model maps input to a fixed-length vector, typically a few hundred to a few thousand dimensions. Similarity is measured by cosine distance or dot product, and vectors are stored in an index for fast nearest-neighbour search.

Why it matters

Embeddings are the connective tissue of modern AI systems: search, RAG, clustering, deduplication, recommendation and classification all run on them.

Common uses

→Semantic search
→RAG retrieval
→Deduplication and clustering
→Recommendation
→Zero-shot classification by nearest label

Strengths

✓Language and synonym aware
✓Cheap to compute and store
✓Works across modalities

Watch for

✓Model-specific — vectors are not interchangeable
✓Re-embedding required when the model changes
✓Poor at exact keyword or identifier matching

Continue exploring

Vector Database Semantic Search Retrieval-Augmented Generation Clustering Multimodal AI