AI gender classification on a legacy database

Problem

The customer profile table lacked a gender field, which became important in the new data model.

Architecture

flowchart LR
  DB[(Legacy database)] --> EXT[Name extraction]
  EXT --> ML[gender_guesser.detector]
  ML --> CLS[Classification ~87%]
  CLS --> UPD[Database update]
  classDef a fill:#0d1525,stroke:#3b82f6,color:#e2e8f0
  classDef g fill:#0d1525,stroke:#10b981,color:#e2e8f0
  class DB,EXT,ML a
  class CLS,UPD g

Solution

• Extracted names from the legacy database.
• Used the gender_guesser.detector library (Python) to infer gender.
• Validated accuracy against a reference corpus.
• Updated the database with the inferred values.

Results

• Backfill with roughly 87% accuracy.
• Gender field available for the new data model.
• Simple, low-cost solution for a data gap.

Stack