ApplicantEvidence Intake, Critérios e Revisão de Candidaturas

O que está incluído

• Vacancy, requisition, location, level and approved version
• Essential, desired, trainable and administrative criteria separated
• Criterion definition, evidence examples and reviewer guidance
• Criterion cannot encode protected trait or proxy without authorized review
• Criteria frozen before application review; amendment creates version
• Candidate notice explains assisted processing and human decision
• Resume/document upload with malware scan and format validation
• Original document hash and immutable received timestamp
• OCR/parser extracts employer, role, dates, education, skills and languages
• Every extracted fact links page/span and parser version
• Unknown, ambiguous and conflicting facts explicit
• Candidate can review/correct configured extracted fields
• No inference of age, gender, race, health, religion, family status or personality
• Address/photo/name masking in blind-review mode
• No facial, voice or emotion analysis
• No social-profile enrichment or scraping
• Application answers separate factual eligibility from assessment
• Knockout criterion only when lawful, job-related and approved
• Administrative incomplete state requests information before decision
• Requirement-evidence matrix with met/not evidenced/unclear, not 0-100 fit score
• Transferable evidence and equivalent experience can be recorded
• Model suggestion never final and reviewer sees source
• Queue priority by received time/SLA or explicit program rule, not predicted value
• Independent second review for configured adverse outcome
• Reviewer decision advance/hold/request info/decline with reason category
• No bulk decline solely from model output
• Override and disagreement retained for calibration
• Candidate communication generated as draft and approved
• Reason explanation uses approved factual category without exposing internal notes
• Correction, reconsideration or accommodation channel
• Duplicate application merge with candidate confirmation policy
• ATS stage/event synchronization and idempotency
• Retention and deletion by candidate, vacancy and legal-hold policy supplied
• Bias/quality monitoring uses lawful aggregate slices and minimum groups
• Metrics extraction accuracy, not-evidenced rate, review time, overrides and outcome flow
• Evaluation by resume layout/language/career pattern and rare criteria
• Pilot for four job families, twenty criteria and two review paths