Semantic search is a search method that does not only match exact words but understands meaning. Instead of a classical index of keywords, query and content are represented as embeddings and compared via vector similarity.
For job search this is a major step forward: searching for "social field, education, working with children" surfaces roles such as nursery teacher, school teacher or family-support worker even if those exact phrases were not used. Synonyms, industry contexts and activity descriptions are taken into account.
In practice, semantic search is often combined with classic keyword filters and geo search. This "hybrid search" couples precision (keywords) with recall (semantics) and then filters hard on region or weekly hours. Re-ranking with an LLM can further improve the top results.
Lunigi uses this approach for the daily job email – candidates phrase wishes in their own words, the system translates them semantically.