The normal route with fission would be to convert the radiation energy into heat, and then convert that into electricity. This is not very efficient. In theory, the nuclear energy has very little entropy, so thermodynamics lets you convert a lot of that energy directly into other forms, but to do that we would have to control the geometry of the nuclear collision, or do some Maxwell's Demon magic to extract the energy from the energetic particles in whatever direction they fly off in. I have always imagined that is what 'dilithium crystals' did. That is way beyond anything we can do at present, even if thermodynamics allows it: we can align nuclear spins at millikelvin temperatures; controlling the geometry of how two tiny balls hit each other is going to be very hard.
If you convert your nuclear energy into heat, you have to get rid of a lot of heat. Not only will you have to get rid of the kilowatts of heat that your apparatus is consuming, the power plant will have to get rid of many times that. It won't be like a battery - it will have to have a lot of heat sinking.
However, suppose that all you want is a burst of radiation. This gets around the efficiency question because the radiation is the end product. As a lightbulb burns out, it will momentarily generate enormous temperatures, as the inductance of the apparatus wants to keep the current flowing, but the last bit of conductor turns straight from solid tungsten to plasma at metallic densities, emitting hard thermal x-rays from temperatures millions of degrees. Suppose your wire had a deuterium-tritium filled void in the centre. If you could control the implosion geometry, you could get fusion. You now have to generate a pulsed current of millions of amps for a few tens of picoseconds to drive the thing. This could be done using a Marx generator, and those can be huge. If you want a one-off device, it could be a slab of piezo crystal and C6 explosive sandwich. The whole gadget could be hand-sized. It might be possible to make a device that worked more than once if you could store the energy from one implosion and use it to drive the next.
That was the answer to the first half of the question. The actual reactor is sub-millimetre scale but the energy to drive the implosion takes some organising.
Next: the radiation. D-T fusion generates a high energy gamma and neutron. Those are very penetrating radiations. It would go through 30 cm of aluminium, or similar alloy. However, it would be stopped by a few millimetres of depleted uranium, which would also release more energy. This is good design from a radiation shielding point of view, but it also means you have in effect made a fast breeder reactor, which is where we are now. Fusion is not 'unlimited clean energy'.
If I more the goalposts a bit, I can get you in a much more feasible place. Suppose you have a Thorium reactor, that does not have by-products that can be used for weapons. It does not need the uranium processing cycle that uranium uses, so you could seal the reactor with its fuel, and when it's power has dropped too far to be useable, you dump it or recycle it. If you are using Uranium, you can seal it if you have a wave reactor - one half conventional and one half fast breeder. But that's another story.
If you are using this to drive a car, you only need the peak power for acceleration. It is possible to have a sporty car with a 300cc engine if you put the energy into a flywheel or an ultra capacitor, and get it out for the bursts when it is necessary. This would not work for a helicopter, which is fighting like mad all the time not to fall, but it would work for many other things. If you have a Thorium reactor providing steady power, and put smarts into your powertrain, you should get a lot more out of a smaller power source. This is the way electric cars are going. If you have a powered exoskeleton, it will have a springy step that recycles the energy from one pace to make the next. It can use technology from the near future.
But probably not fusion IMHO.