High DC gain with short channel MOSFETs is going to
be tough. Have you been playing with common gate,
stacked FETs to get longer "effective L"? If not then
I'd expect you to struggle with AVOL.
And of course there's the questionable usefulness of
such low voltage op amps for anything.
At 0.9V I'd be thinking of making any analog functions
current-mode. But I know a lot of people are really
determined to put lousy little voltage mode (usually OTA)
op amps in their SoC library. I work with one such bunch.
From what I've seen they are using the usual architectures
and living with pretty poor (by piece-part standards)
results. Just so they can say they've got one. Or N
variants (all of them lousy, but slightly different in how
much and what's worst).
900mV range and 10mV input error rollup would make it a
sub-7-bit accuracy proposition. You want to spend time on
such a world-beater "product", go right on ahead.
But you might find it worthwhile to first imagine or find an
application for such an op amp (I can't, but don't let that
stop you) and let it provide you the "care-abouts" (specs
or more like a subset, of). That would then leave you only
the problem of execution.