AADHAAR_Extractor: Indian PII Fine-Tune (fastino/gliner2-base-v1)

This is a fine-tuned version of the fastino/gliner2-base-v1 architecture, optimized specifically for extracting Indian Personally Identifiable Information (PII) from unstructured text.

The model was trained to replace brittle RegEx pipelines with a generalized neural extractor. It identifies complex identity and financial markers regardless of surrounding sentence structure.

Supported Entities

The model is trained to predict and extract the following exact labels:

• Person Name
• PAN Number
• Aadhaar Number
• IFSC Code
• Bank Name

Model Details

• Base Architecture: fastino/gliner2-base-v1
• Task: Named Entity Recognition (NER) / PII Redaction
• Training Data: Synthetically generated records mirroring real-world Indian financial and identity document formats.
• Language: English (with Indian contextual formats)

Usage (Inference)

Requires the gliner library. Ensure your environment has the required dependencies installed.

pip install gliner2