I recently learned about algorithms called Learn-to-Rank (LTR) from a talk that Doug Turnbell gave (he works at Reddit). I actually didn’t know they existed. “LTR models are specifically designed for ranking tasks, optimizing metrics like NDCG or MAP, which are more aligned with ranking quality.” Also learned that xgboost has an implementation of LTR here: https://xgboost.readthedocs.io/en/stable/tutorials/learning_to_rank.html.
I then had a conversation with ChatGPT on how this could be applied to Candidate Match. (Side note: I find LLMs to be fantastic for this type of thinking/research, I copy pasted the xgboost LTR documentation into the prompt for context).