Issue in multiple documents embeddings

I have created embeddings for multiple documents, such as the resumes of three candidates. However, when I query the documents, the returned content is mixed — for example, projects from Candidate 1 appear in the summary of Candidate 2.

How can I resolve this issue without using a vector database?

I cannot rely on candidate IDs, as the use case is not limited to resumes — the documents could also be history or chemistry books.