We are more keen on minimal designs rather than complex ones, however -depending on the concept adopted - some complexity and detail is allowed (and in some cases encouraged).
Below we give some descriptions/abstractions about a Machine Learning/data analytics system. This is just for some inspiration, don't feel restricted by this, feel free to improvise.
Any machine learning/data analytics system deals with an input into the system and an output from the system.
On the one side (could be left, or top, or in an angle - whatever) is the input which comprise the unstructured/messy data whereas on the opposite side there are the processed/structured information. In essence of the data have been distilled - their deeper meaning has been understood.
The limits between the two sides could be abrupt or gradual - could potentially both work.
An example (don't take it as restrictive) is shown in the attached image. A stream of unstructured data enters the system, which is shown, for example, by a cube and then the data are getting structured and useful information gets out of the system.