embedding check

The informal act of running a quick sanity test on your vector embeddings to verify they're actually capturing semantic meaning correctly — kind of like a vibe check but for your ML pipeline. An embedding check might involve querying similar words to see if the model clusters them sensibly, or confirming that cosine similarity scores feel right. It's the unglamorous but essential quality-assurance step that ML engineers do before trusting their retrieval or recommendation systems to work properly.