Why invert the BJT? It works just fine in the normal way, and to much higher voltages.
Note that a speed-up cap (say 47pF across R2) may prove beneficial. The input drive has to be symmetrical, which is usually the case (CMOS driving pin).
I've used a similar mechanism in other circuits before; here's a complementary 12V level shifter for example:
https://www.seventransistorlabs.com/Class_D_Amp.pdfSpecifically, after IC1B; note that its output only pulls up to 2*Vbe, and it can pull down towards zero (at a few mA). When pulling down, C2 quickly shunts off Q3, and D3+D4 biases Q4 on. Q4 turns off a bit slower (it does go into saturation) but Q3 only turns on after Q4's storage time is done, so the timing is fairly close (within 20ns I think it was). The pull-ups are CCS for speed, but they're not plain old CCS; in fact as one side saturates, base bias is shunted (D6-D7 drop less then 2*Vbe) saving bias current into the low side. More specifically, it acts as a current mirror, so the wasted bias current will about equal the load current. (Which as this is a gate driver, DC bias is about zero, so it's pretty efficient.)
When a similar cross-coupling drive is used with PMOS, you have a very similar circuit as used in actual CMOS level shifters. (The difference there is, the PMOS Rds(on) is made relatively large so they act as conditional pull-ups that the NMOS can still pull down, and the gate and drain are cross connected so they act as a flip-flop, which removes static current draw.)
The main trouble with discrete MOS is -- besides the obligatory* body diode -- the capacitance is just so high. They're all made for switching modest loads, not light logic signals. Mostly super old types too (like BSS138 and 2N7002). And RF transistors don't offer anything, as they've been cycled through regularly as technology and applications advance; you can get a 60GHz transistor today, but nothing for just a few.
*There are very few devices available with separate substrate pin anymore; mostly for odd edge cases like battery switching, so rated for much more current (higher capacitance) than is useful for logic signals.
So if you do need speed, and can't afford the lower pull-up resistors and everything, BJTs may be an option. You do need the extra resistor(s) though.
Tim