Research / Paper

Deduplication and Probabilistic Record Linkage

Comparative study of Transformer-based, Gradient Boosted, and Fellegi-Sunter models for identity resolution.

Built during a summer research fellowship, this project combines record linkage, evaluation design, and applied identity reasoning.

1st Place at Mid-Hudson Valley Tech MeetPaper + repo

Overview

Comparative study of Transformer-based, Gradient Boosted, and Fellegi-Sunter models for identity resolution.

A comparative record-linkage study using transformer, tree-based, and Fellegi-Sunter approaches.

The composite scoring protocol outperformed the single-architecture approaches and supported the award-winning presentation.