Data-Centric

An AI engineering paradigm that emphasizes systematically improving data quality rather than adjusting model architecture, including finding annotation errors, supplementing edge cases, and balancing data distributions.