.svg)
What is Data-Centric AI?
Data-centric AI is the practice of systematically improving datasets—labels, coverage, and consistency—rather than only tuning model architectures, on the premise that data quality often dominates performance gains.
It pairs iterative data curation with fixed evaluation benchmarks so each labeling or cleaning change is measured, not assumed.
Related reading on Giskard
Iterate on data with evaluation feedback — Use failing cases from Giskard suites as the next labeling and curation backlog—data-centric AI needs measurable quality gates. Learn more: Announcing Giskard OSS v3.
Authoritative reference: Data-Centric AI
