A deep learning-based python package for heterogeneous data integration is provided. The major functionality of our package is to integrate any numerical data generated from multiple domain regardless of time series or non-time series. The most significant advantage of this package is the flexibility in which irregular time series data can be processed. The package can incorporate any number of modalities by which combining multiple GRUs with simple concatenation-based vector integration. In fields where available samples are small and varied depending on modalities, our package can utilize every available samples including non-overlapping samples as well as overlapping samples.