Cross-Platform Entity Resolution: Deduplicating SaaS Identities

In enterprise data engineering, a major challenge when constructing a corporate knowledge graph is identity fragmentation.

A single senior engineer might exist across SaaS applications as:

• @alex_sf in Slack
• alex-chen-code on GitHub
• [email protected] in Jira
• Alex Chen in Zoom meeting transcripts

If an enterprise AI knowledge graph treats these representations as four separate people, search queries cannot accurately aggregate contributor context or trace decisions across systems.

In this systems engineering breakdown, we explore how Memora executes Cross-Platform Entity Resolution to merge fragmented user profiles into unified contributor nodes.

3-Layer Identity Resolution Pipeline

┌─────────────────────────────────────────────────────────────────────────┐
│                 3-LAYER IDENTITY RESOLUTION PIPELINE                    │
├─────────────────────────────────────────────────────────────────────────┤
│ Layer 1: Deterministic Email & OAuth ID Match (Okta / Azure AD Sync)   │
├─────────────────────────────────────────────────────────────────────────┤
│ Layer 2: Probabilistic String Similarity (Jaro-Winkler & Levenshtein)   │
├─────────────────────────────────────────────────────────────────────────┤
│ Layer 3: Co-Occurrence Graph Topology Clustering                       │
└─────────────────────────────────────────────────────────────────────────┘

Layer 1: Deterministic Email & SSO Matching

Memora queries enterprise Identity Providers (IdPs) via SAML / SCIM APIs. If a user's GitHub account and Jira account share the same corporate SSO email ([email protected]), nodes are merged automatically with 100% confidence.

Layer 2: Probabilistic Name Similarity

When email handles are obscured (e.g., matching a Slack username to a Git commit author name), Memora calculates Jaro-Winkler string similarity:

def jaro_winkler_similarity(s1: str, s2: str) -> float:
    # Measures string edit distance weighted by matching prefix
    pass

Layer 3: Co-Occurrence Graph Topology Clustering

If two user nodes frequently co-occur in the same pull requests and Slack threads, graph clustering algorithms increase their identity merge probability score.

Related Technical Resources

• How AI Builds Organizational Memory
• Entity Extraction Schemas for Technical Codebases
• Graph RAG vs Vector RAG Comparison