Data Mesh versus Central Warehouses for Trustworthy AI Pipelines: A Reproducibility-Centred Evaluation

Abstract

Data-mesh and central-warehouse architectures allocate ownership and governance responsibilities in different ways. The objective is to compare their implications for provenance, access control and reproducible AI training across enterprise units. Relevant methodological and data-quality literature informs the design of the sampling and validation strategy [3–7]. The proposed method is a design-science comparison using matched data-product scenarios, architecture diagrams and traceability tests. Comparators will include domain-owned data products, a central warehouse and a hybrid governance model. Related enterprise-storage studies inform the assessment of operational reliability [1] and the deployment assumptions adopted for system-level comparison [2]. The reproducibility extension records dataset versions, model settings, exclusion rules and provenance before any held-out evaluation. Statistical and computational sources help define the reproducibility checks and comparison procedures [8–12]. The primary outcomes are dataset lineage completeness, policy violations, integration effort and model retraining reproducibility. Potential biases include conflicting data contracts, institutional access rules, ownership disputes and cross-domain schema changes. An independent replication exercise would repeat the main analysis from the documented artefacts and quantify disagreements. Additional research informs the prespecified stress tests and limits on interpretation [13–17]. The intended contribution is an auditable research protocol rather than an unsupported claim of effectiveness; related work frames its transferability limits [18–23].