migrateforce/resume-parser

Summary

This skill extracts structured, machine-readable data from unstructured resumes. It handles multiple formats (PDF, DOCX, plain text, LinkedIn exports) and normalizes the output for downstream processing like candidate screening or database storage.

Inputs

| Field | Type | Required | Description | |-------|------|----------|-------------| | resume_content | text/file | yes | Raw resume content or file path | | format | enum | no | pdf, docx, txt, linkedin (auto-detected if not specified) | | extraction_level | enum | no | basic, standard, detailed (default: standard) |

Outputs

| Field | Type | Description | |-------|------|-------------| | candidate_name | string | Full name | | email | string | Primary email address | | phone | string | Phone number (normalized) | | location | object | City, state, country | | linkedin_url | string | LinkedIn profile URL | | summary | string | Professional summary/objective | | experience | array | Work history with dates, titles, companies, descriptions | | education | array | Degrees, institutions, dates, GPA | | skills | array | Technical and soft skills | | certifications | array | Professional certifications | | languages | array | Languages and proficiency levels | | parse_confidence | number | 0-100 confidence score |

Workflow

1. Format Detection

• Identify input format (PDF, DOCX, TXT, LinkedIn JSON)
• Extract raw text using appropriate parser
• Preserve section structure where possible

2. Section Identification

Identify standard resume sections:

• Contact Information
• Summary/Objective
• Work Experience
• Education
• Skills
• Certifications
• Projects
• Publications

3. Entity Extraction

Contact Info:

• Name (handle international name formats)
• Email (validate format)
• Phone (normalize to E.164)
• Location (city, state, country)
• LinkedIn URL

Experience:

{
  "company": "Acme Corp",
  "title": "Senior Engineer",
  "start_date": "2020-03",
  "end_date": "2023-06",
  "current": false,
  "description": "Led backend team...",
  "achievements": ["Reduced latency 40%", "Mentored 3 juniors"]
}

Education:

{
  "institution": "MIT",
  "degree": "BS",
  "field": "Computer Science",
  "graduation_date": "2018-05",
  "gpa": 3.8
}

4. Skill Normalization

• Map variations to canonical names (JS → JavaScript)
• Categorize: languages, frameworks, tools, soft skills
• Infer skills from experience descriptions

5. Quality Scoring

Calculate parse confidence based on:

• Fields successfully extracted
• Date format consistency
• Section completeness

Supported Formats

| Format | Support Level | Notes | |--------|--------------|-------| | PDF (text) | Full | Standard text-based PDFs | | PDF (scanned) | Partial | Requires OCR, lower accuracy | | DOCX | Full | Microsoft Word documents | | DOC | Full | Legacy Word format | | TXT | Full | Plain text | | LinkedIn JSON | Full | LinkedIn data export | | HTML | Full | Web-based resumes |

Example Output

json
{
  "candidate_name": "Sarah Chen",
  "email": "sarah.chen@email.com",
  "phone": "+1-415-555-0123",
  "location": {
    "city": "San Francisco",
    "state": "CA",
    "country": "US"
  },
  "linkedin_url": "linkedin.com/in/sarahchen",
  "summary": "Senior software engineer with 8 years of experience...",
  "experience": [
    {
      "company": "Stripe",
      "title": "Staff Engineer",
      "start_date": "2021-01",
      "end_date": null,
      "current": true,
      "description": "Lead payments infrastructure team",
      "achievements": [
        "Scaled payment processing 10x",
        "Reduced fraud by 30%"
      ]
    }
  ],
  "education": [
    {
      "institution": "Stanford University",
      "degree": "MS",
      "field": "Computer Science",
      "graduation_date": "2015-06"
    }
  ],
  "skills": {
    "languages": ["Python", "Go", "TypeScript"],
    "frameworks": ["Django", "React", "Kubernetes"],
    "tools": ["PostgreSQL", "Redis", "AWS"],
    "soft_skills": ["Leadership", "Mentoring"]
  },
  "certifications": [
    {
      "name": "AWS Solutions Architect",
      "issuer": "Amazon",
      "date": "2022-03"
    }
  ],
  "parse_confidence": 95
}

Error Handling

| Error | Handling | |-------|----------| | Unreadable PDF | Return error with suggestion to re-upload | | Missing contact info | Flag as incomplete, extract what's available | | Ambiguous dates | Use best guess, flag for review | | Non-English resume | Attempt extraction, note language |

Integration Points

• Upstream: File uploads, email attachments, ATS imports
• Downstream: candidate-screening, database storage, search indexing