Golden Record

Designed and built a backend identity-resolution system from the ground up to help unify fragmented customer records, identify duplicate entities, and manage relationships across approximately 23,000 CRM records.

The system combines deterministic and probabilistic matching techniques to identify records that belong to the same individual, even when names, contact information, and other attributes differ or contain inconsistencies. It also supports record merge and unmerge operations, along with household relationship management, helping maintain a more complete and accurate view of customer identities.

I developed a multi-stage matching pipeline using Node.js, TypeScript, PostgreSQL, and Google Cloud, incorporating eligibility rules, blocking strategies, and structured matching paths to improve efficiency and reduce unnecessary comparisons.

Designed and implemented the identity-resolution backend and matching engine from scratch.

Developed deterministic and probabilistic record-matching strategies.

Implemented eligibility gates and blocking strategies to narrow candidate comparisons.

Built structured matching paths to handle different levels of available data.

Implemented record merge and unmerge capabilities to support controlled identity consolidation and correction.

Developed household relationship management to connect related individuals and maintain household-level context.

Implemented database-level tenant isolation and access controls.

Added audit logging and correlation IDs for traceability and debugging.

Developed an evaluation harness to assess matching behavior and improve accuracy.

Technology: Node.js, TypeScript, PostgreSQL, Google Cloud Platform (GCP).

Outcome: Built a complete identity-resolution system with matching, record lifecycle management, and household relationship capabilities—providing a structured, auditable approach to resolving duplicate identities and organizing customer data.