← Projects

Address Matching & Validation

2024 – Present · Multiple enterprise clients

Architected a configurable address-matching pipeline for Indian addresses at scale—normalize, match, score—reducing manual verification and feasibility checks for enterprise clients.

PythonPostgreSQLFastAPIDockerLinux

System design

Modular pipeline: ingest → normalize (script, abbreviations, casing) → match (exact/block then fuzzy) → score → output. Configurable rules and thresholds per client; observability on match rates and manual review samples for tuning.

Highlights

  • Led system design and implementation of the address-matching pipeline; defined interfaces and deployment strategy.
  • Significantly reduced manual verification and feasibility effort; pipeline integrates into multiple client workflows with clear contracts.
  • Designed for extensibility: swap normalization and reference data for other geographies without changing core flow.