how many boards do you need to make in prototyping, and for real? How developed is your board - how confident are you that this should work?
Paging @Justin who has a picknplace that may not get much use??
Personally, I’d always do prototype boards myself, but I’m a tightar$e. 50 is not that much - I’ve modded a toaster oven for reflow, and I have an SMT hot-air gun (not a rework station) which is good enough for most things needed, except fine pitch, small amounts of solder etc… then I’m screwed.
The cost in doing small run boards is the setup time on a machine - the NRE, non-recurring engineering time - making the pnp know how to grab the component, and put it in the right spot/orientation. Doing that takes X minutes/hours, and they then have to recoup that cost spread out over your job - which in a small run prototype could be as little as say 10 boards, making the per-board cost very high. Since it’s a prototype, that NRE is wasted unless your board was perfect, then they need to have minimal setup time for doing a production run. So your challenge is making the NRE actually NR, and you can do that through spending money (with them to do tests) or through labour doing the assembly yourself the first time