Candidate embeddings

A candidate embedding is a numerical representation of a profile's content that captures what it means, not the exact words it uses. Two profiles that describe similar work end up represented similarly, even if they share almost no vocabulary — which is what lets a search find "led a small engineering team" as a match for "managed a team of four engineers."

You don't need the underlying math to benefit from it: the practical consequence is that writing your candidate profile the way you'd describe your work to a person is exactly what produces a strong match, because that's what the representation is built to capture.

This is the mechanism behind semantic matching. For the full picture, see how AI job matching works.