The 2N2222A isn't really up to switching task you're asking it to do... there's 3 key areas it's failing in:
1. The (maximum) collector current rating for the 2N2222A seems to vary all over the place depending upon manufacturer/package etc (a quick Google around revealed values from 600 mA - 1A). Depending on your variant, you might be exceeding the 'absolute maximum' spec for your part, which is a very bad thing to do
More importantly, though:
2. The "DC current gain" family of curves in the datasheet are a useful guide* as to the drive requirements for a BJT. From the data sheet I found at
https://users.ece.utexas.edu/~valvano/Datasheets/2N2222.pdf, Figure 3 (with some generous extrapolation) suggests that the current gain of the transistor could be expected to be as poor as ~20 at your desired 750 mA load current. That means you'd need to provide it with 750/20 ~= 37 mA of base current for it to enter saturation. With the circuit you have shown, the maximum possible base current is: 3.3V (from the Xbee) - 0.7V (Vbe drop) / 1K = 2.6 mA, over an order of magnitude too small! As such, the transistor can't turn on fully, and is dissipating large amounts of heat.
3. How much heat? The transistor dissipation = collector current x Vce voltage. Figure 4 shows the sort of Vce you'd expect for a given base current/collector current combination. At 2.6 mA base current, Vce is off the chart (> 1 V) for collector current somewhere between 150 & 500 mA. This tells us that the transistor a) won't be able to sink the required 750 mA, and b) if it did, it has many VOLTS of drop between collector and emitter. Thus the package dissipation is likely to be in the order of WATTS, and far in excess of the specified absolute maximum ratings. I have no doubt it gets hot!
So - what to do?
Easy. Get a bigger (+better) transistor. If you like BJT's, then just beware that bigger devices often have poorer current gains - an elegant solution in your application is the so called "darlington" configuration (Google it). You can get them as single devices (eg. BD681) too and such a part would happily run from your 1K base resistor.
Alternatively, you can use a power MOSFET - no 1K resistor needed (although a pull-down of similar value isn't a bad safety feature to add) and just about anything in a TO-220 package will have enough grunt for you. Just beware their threshold voltage spec - at 3.3V you might be a little marginal for some devices. (Check the drain current vs Vgs specifications/plots).
Good luck!
* Note: I said GUIDE. The DC current gain of a transistor is an extremely poorly controlled device parameter - and strongly temperature sensitive - and should NEVER be used to design bias networks. Use them for sanity checking base current drive requirements ONLY!