I hadn't seen the home made guy's stuff, it's an interesting take on the problem. But I think he realised that he was a bit over his head. Trouble is that the ideas are really simple, but usually the problem with these optical ideas are a) you have to machine things at ridiculous tolerances and b) when you have a spot of 1um, vibration, humidity and temperature actually can change things a lot (your metal frames expand and contract, diode changes its wavelength with temp, etc). And yes, I have literally watched a diode undergo a wavelength shift as it warmed up (using an optical spectrum analyser).
I don't know about "good" home made stuff (although you can see a few ok-ish videos on youtube), but you can definitely do it well in a commercial setting. I work with lasers, so I find it interesting as an engineering problem, but I don't really see it as a real competitor to fab houses. I think it might be popular in MEMS research for prototyping.
I think the best option nowadays is to get a 405nm diode designed for a blu-ray drive. I believe DuPont make a 405nm sensitive photoresist that you can use with it. The key is to get good focussing to reduce the spot size and decrease the required exposure time. Assuming you had the right mounting gear in a lab setting, getting a spot size of a few micron is definitely achievable. If you want to do the stage mechanics at this sort of level, you will probably need something with decent (aka precision THK) ballscrews with not just a rotary-encoder style servo, but with linear encoders to counteract any backlash. Or possibly linear motors from a pick-n-place. You could do it with galvanometers too, but you would make the focussing problem even more sensitive.
The true UV lasers I have seen in the past (<400nm) have cost about as much as a car. So I wouldn't go down that road unless you have a lot of disposable income
Even just the properly certified safety glasses for this stuff are a few hundred dollars.