Combining text-embeddings and tabular data for human ressources

I want to build an ai-system that helps recruiters to find the perfect match on candidate-profiles and jobdescriptions on an internal database.

I started to generate text-embeddings from both items and use a vectordatabase to find similar profiles. This works quite wells but cannot take all the parameters into account.

E.g. you find a match between a candidate from India and a job description on-site in Germany. The candidate won’t resettle to Germany.

Therefore I want to build a ML-Modell that generates matching suggestions based on tabular data.

Do you have any suggestion on an ML-Algorithm that works for this problem? I can create a trainingdataset with about 500 matches between jobdescription and candidate profile.

Next question is: how can I aggregate the outputs from both models into a shortlist of candidate suggestions?