Different neural network designs for different inputs

How do I make a different design for different inputs? For instance, let say I have 10 features and want to build a feedforward network for five inputs and LSTMs for the five other inputs, how do I implement that? I assume that the outputs will be aggregated at the end, though a linear node for instance

Take a look on the Keras Functional API: The Functional API  |  TensorFlow Core

You can do exactly what you want and on the link there are some comments on that ( Manipulate complex graph topologies)