Scope (Crawler + ML)

I'm currently trying to define the scope of a proyect of mine, trying to
resume it in a couple of use cases. The idea is, more or less, as follows:
imagine a "search engine" or a kind of "crawler" which can "index" diverse
backends, services and APIs. This "engine" shall allow for the interaction
with what was indexed without needing to "leave the page" (by means of
layers of abstraction).

I mean, you could "search", for example, some car for make and model (data,
type, attributes), filter and sort for those attributes, navigate
relationships of attributes and entity occurrences: "some guy sold a
similar car to some other guy on some date with some comments / opinions".
Or filtering through facets / roles: a car of some make / model in some
location by publishing date. And, eventually, perform some operation as
bidding some amount for the car.

It's funny but, once an adequate metadata representation (integrated from
sources / APIs) in an appropriate format (features encoding) for being able
to use Machine Learning, this three use cases seems to correspond with /
may be addressed with "canonical" ML problems: classification, clustering
and regression (being this networks "embedded" into an integration /
abstraction framework.

Any opinions are welcome.

Regards,
Sebastián.