← 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.