Is there a way to build a multi-modal model in keras with input from TFRecords?

Question

I have built the model below to take both an image input and some tabular data input into a keras model. The architecture is based on mobilenetv2 and implemented in keras:

<