1) Inputs

From the resume:

• latest_experience.title → short title string (e.g., “Data Scientist”).

• responsibilities → text describing all work.

• resume_ie.technologies, industry_tags, experience.years_overall,

  seniority_level, primary_families_ranked (optional), education, location/countries.

From each listing (pre):

• e_title, e_text → precomputed embeddings for job title and job text.
• llm_ie → LLM extraction (required/proferred tech, industry tags, job_family, seniority, region_restrictions, etc.).
• required_education_inferred, region_restrictions_inferred → one-time heuristics to catch PhD/region hints in title/raw text.

2) Hard filter (pass/fail before scoring)

A listing is discarded unless it passes all three:

Years of experience

• need = llm_ie.required_experience.years_min
• have = resume_ie.experience.years_overall
• If either is None ⇒ don’t block.
• Else require have ≥ need.

Education

• Build required_edu = required_education_inferred (fallback: llm_ie.required_education).

• If it says “PhD” (or Master’s/Bachelor’s), resume must mention that degree (string contains, case-insensitive).

  If no required degree is present, don’t block.

Location/Region