Built graph-based identity resolution engine processing millions of records with 95% match accuracy

The Challenge

Breach data investigations required manually connecting scattered identity fragments across millions of records. Analysts spent hours tracing relationships between emails, phones, usernames, and IDs with high false positive rates from common values like "123456" passwords.

What I Did

→ Designed multi-signal matching with weighted confidence

→ Built recursive graph traversal with configurable depth and breadth limits

→ Implemented dynamic supernode detection to filter high-frequency noise

→ Integrated Elasticsearch + Neo4j for search and visualization

Results

95% confidence accuracy on 4-level recursive search, Sub-30 second query response on 1M+ records