data-versioning.md

Data Versioning

This document describes the data versioning strategy for the MLOps project.

Approach

For this learning project, we'll use a simplified data versioning approach:

1. Training Data: Stored in data/ directory (not committed to git)
2. Data Loading: Scripts in src/data/ handle data loading
3. Data Preprocessing: Preprocessing steps are versioned as code
4. Data Validation: Basic validation to ensure data quality

Future Enhancements

As the project evolves, we may add:

• DVC (Data Version Control) for data versioning
• Data validation schemas
• Data lineage tracking
• Automated data quality checks