It's hard to see why they couldn't.
While it's important to remember that we don't actually know much about how intelligence evolves (we only have one instance to extrapolate from!), we think we have some pretty good guesses which seem consistant with the rest of evolutionary biology.
It seems likely that intelligence evolves most readily in social creatures where reproduction is most successful if you can dominate a social structure without injuring yourself. So brains (and, later, a golden tongue) are adaptive. This creates a more complex society, driving further development of brains and language. (Twitter, of course, reverses this process.) One biologist commented that language evolved so we could lie to each other better. Smart is successful, and successful is sexy.
I can see no reason this couldn't happen in marsupials just as well as it has in mammals. The key step would be the development of a species with primate-like social groups.
(Obviously there's more to the development of intelligence than that, since primate-like social groups probably are not unique to primates, but no one knows the details, though there are lots of theories and lots of popular books talking about it. (One common idea is that intelligence requires having hands and being capable of fine manipulation.) Still, nothing that would make it less likely in marsupials.)
As to the specific question of kangaroos, probably not. Kangaroos are grazing animals who live in herds and while there have been and are lots of species like that, none of them are ranked high in intelligence. But it appears to be possible for kangaroos to evolve into a more social species which then evolves intelligence.
As to what they would look like, that depends a lot on how quickly intelligence evolved. The physical shape is overwhelmingly determined by a species' lifestyle. A grazing animal today quickly evolving intelligence would probably still be a grazing animal and continue to be adapted to grazing, etc. There's very little reason to think that intelligence goes along with only one or a small set of body plans. (Larger heads, certainly; Manipulators of some sort (like hands) probable. Not much else.)