Job recommendation system

Hi, I am trying to build a job recommendation system based on each resume submitted, I am thinking to use LLMs to do so, but does anyone have any suggestions on how to use LLMs to do so? Thank you.

1 Like

Just talked about this yesterday over here

Hi, I just read this thread, would you suggest extracting each segment out (education, work experiences, skills) individually and embed each segment, assigning different weightages to calculate the cosine similarity score for each segment, using the sum as the total score?

If your goal is to recommend a set of jobs that are matched to an uploaded resume. I would correlate the resume, to labeled resumes (ground truth) and correlation is done via embeddings/keywords. The label is the job, and you grab the most correlated jobs, or top list of matching jobs as your result.

There are other thoughts on that thread on how to do this, and other things. For example @Diet has a more detailed/sophisticated approach by weighting embeddings and hours of experience to arrive at an overall score.

So it depends, but try the easier things first, and evolve from there.

1 Like