you definitely can, as long long you can have a libname assigned to the Teradata source.
About optimization, it's nearly impossible to have an opinion with this little information.
Probably will processing be slower than having it all happen in Teradata using pass-thru. So the question is, is there any functionality with the Look-up transform you need that can't be handled by a SQL join?
I think you should wait until you can get a hand on real data and try it in real life.
Data never sleeps