Back to overview

How It Works

A technical overview of the verification process, from data ingestion to output artifacts.

System Overview

Vericor operates as a verification layer that sits between your data sources (ATS, candidate submissions, third-party databases) and your decision-making processes. The system processes candidate information through four distinct stages, producing structured confidence signals with full evidence trails.

Architecture

Your Systems (ATS, HRIS, Forms) → Vericor APIVerification EngineOutput ArtifactsYour Systems

Vericor does not store candidate data beyond the verification session unless explicitly configured for audit retention. All processing occurs in isolated, customer-specific environments.

Verification Stages

Stage 1

Data Ingestion

Candidate information is received via API, file upload, or direct integration. The system normalizes data into a structured format, extracting key claims: employment dates, job titles, credentials, skills, and education.

Processing: Deterministic parsing and normalization. No AI is used at this stage.

Stage 2

Multi-Source Verification

Each claim is verified against available sources: public records, credential databases, employment verification services, and structured web data. The system cross-references multiple sources where available.

Processing: Deterministic matching and lookup. AI is used for entity resolution when exact matches are unavailable (e.g., company name variations, credential title normalization).

Stage 3

Confidence Modeling

Verification results are aggregated into confidence scores. Each score reflects the strength of evidence, number of corroborating sources, and any discrepancies found. Scores are categorical (High, Moderate, Low, Insufficient Data), not numeric rankings.

Processing: AI is used to synthesize evidence from multiple sources and generate natural-language explanations. The scoring logic itself is deterministic based on evidence thresholds.

Stage 4

Output Artifacts

The system produces structured outputs: confidence signals per claim, an overall verification summary, evidence citations, and optional embeddable badges. All outputs include timestamps and can be exported for audit purposes.

Outputs: JSON API response, PDF report (optional), webhook notification, embeddable verification badge.

Role of AI in the System

AI is one component of the verification system, not the system itself. We use AI selectively where it provides value that deterministic logic cannot:

Where AI Is Used

  • • Entity resolution (matching company name variations)
  • • Credential title normalization across different naming conventions
  • • Synthesizing evidence into human-readable explanations
  • • Identifying potential discrepancies that warrant human review

Where AI Is Not Used

  • • Data parsing and normalization
  • • Database lookups and record matching
  • • Confidence score calculation (threshold-based)
  • • Any form of candidate ranking or recommendation