Designing correct data integration systems is crucial, especially in AI projects where the datasets for analysis must be consistent, homogeneous, up-to-date, representative, and free from significant errors or biases. This is especially true in complex domains such as genomics, where integrating and querying many data sources with different data schemas, quality, and formats is required. This work presents a framework for creating an Automated Virtual Data Integration System combining different solutions presented in the literature and adapted to the particularities of genomic data.