What do you mean by "Servler-less mechanism for video encoding"?
Would it not make sense to encode using CPU power of node providers?
Also, how will you address delivery speed without a CDN?

Other questions are: what formats will you be supporting (HLS, Smooth Streaming, etc...)
And what variants will you be producing? Depending on variant numbers and quality, you might need lots of storage on different nodes. Unless you wanted to encode on the fly (server-less?), which would result in slow streaming of video. Maybe a way around this is to develop a CDN solution as well that stores the variants once they are initially produced and delivered so they do not have to be encoded again.

Some really interesting problems to have to solve.