Resume Fit Scoring Across Four Dimensions

{
  "model": "jev-latest",
  "state": {
    "role": "{{role-title}}",
    "must_haves": "{{must-have-skills}}",
    "resume": "{{resume-text}}"
  },
  "questions": {
    "core_skill_depth": {
      "type": "score",
      "instructions": "How much depth does this candidate show in {{core-skill}}, based only on the resume?",
      "criteria": [
        "Not mentioned",
        "Listed as a skill, no evidence of use",
        "Used in at least one described project or role",
        "Used across several roles with named outcomes",
        "Deep expertise: architecture, performance, or teaching others"
      ]
    },
    "leadership": {
      "type": "score",
      "instructions": "How much experience does this candidate have leading or managing people?",
      "criteria": [
        "No leadership mentioned",
        "Mentored or led a small project informally",
        "Formally managed one team",
        "Managed multiple teams or a department"
      ]
    },
    "domain_match": {
      "type": "score",
      "instructions": "How closely does the candidate's industry experience match {{industry}}?",
      "criteria": [
        "Different industry entirely",
        "Adjacent industry with transferable context",
        "Same industry, different function",
        "Same industry and same function"
      ]
    },
    "generalist": {
      "type": "score",
      "instructions": "How much evidence is there that this candidate picks up unfamiliar tools, roles, or domains outside their core specialty?",
      "criteria": [
        "Only one domain or role mentioned",
        "One instance of switching context",
        "Repeated instances across roles",
        "Track record of ramping up in unfamiliar areas and delivering"
      ]
    },
    "has_must_haves": {
      "type": "noul",
      "instructions": "The resume shows every must-have skill listed in the state, not just some of them"
    }
  }
}

Combine in code, never inside the model. Normalise each score by dividing by its top level (4, 3, 3, 3). Then weight per role: for a senior individual contributor use 0.45 core_skill_depth, 0.10 leadership, 0.25 domain_match, 0.20 generalist; for a manager use 0.20, 0.45, 0.15, 0.20. Reject before scoring when has_must_haves is below 0.5. Keep the four raw scores visible on every ranked candidate so a reviewer can see why the total landed where it did.