Help with an NLP project

I am helping a friend with his project. He wants to design an automated resume filtering model. We are stuck at one point. The flow is as follows:

  1. We identify the various segments in the resume (instead of simply extracting text). For this, we can use a combination of things like font, positioning, keywords, and spaces and so on.
  2. Once it is split, we use named entity recognition to label our data (institutions, locations, etc) We also use a regular expression to convert all the dates into a general format
  3. Finally, we want to extract a score out of each section (skills, work experience, and education). We will decide the weightage of each later but for now, we have a way of scoring skills and education. We are unsure about how to move forward with work exp though. Do you have any ideas for it? The parameters that would be reliable to judge a candidate.

We’ve read 3-4 research papers but nothing seems convincing.

Open to any feedback or good ideas regarding anything about the project.