Cross-domain data integration remains a critical bottleneck in semiconductor sensor development because siloed data sources, inconsistent semantics, and missing alignment mechanisms restrict unified analysis. We present a general-purpose data integration framework that creates a semantic layer and can be applied across industrial contexts. In this work we applied it to Micro-Electro-Mechanical System (MEMS) development by constructing a domain-specific Knowledge Graph (KG) ecosystem that integrates heterogeneous data from the engineering and manufacturing domains. The KG extends the existing infrastructure and uses a modular and expert-aligned ontology to harmonize formats and units and resolve structural inconsistencies. A lightweight enrichment layer infers additional relations, including temporal process sequences and spatial chip relations. SHACL validation enforces standards and flags missing or inconsistent data. To improve accessibility for non-technical users, we provide a natural-language interface for querying the KG (SPARQL-NLI) that benefits from the KG’s unified semantics. The system enables complex cross-domain queries that raw relational data cannot support efficiently. We evaluated the system through competency question coverage, expert feedback, and graph-level analytics. The results show improved data accessibility and more reliable cross-domain querying, although challenges remain in scaling and maintaining the KG. This work demonstrates how a focused semantic integration layer embedded in the existing infrastructure can bridge knowledge gaps and provide reusable and explainable analytics in industrial environments. This supports stakeholders in accessing and exploring cross-domain data without facing common integration hurdles.