I am putting together my own UV box right now and considered what you suggesting. However, I opted for my first design to use a proved type. I am interested in what you might find out.
Will the LED's act like point source lights and cause sharply defined illumination areas? This may be sorted out I suppose by ensuring "sufficient" stand off between light source and image area.
I remember reading that above 375 nm there were increasing losses due to absorption by the cover glass of the printing frame so perhaps unless you are using a vacuum frame this may be of concern to you.
I think it is a wonderful idea to try. The LED efficiency is legendary and would surely be lighter. I wonder about the economics of construction. What is the pricing like? They should far outlast the fluorescent tubes we are using and the heat factor should just about be gone.
Have you thought of building a simply 5 x 7 or 8 x 10 size device and testing?
If you got for please let us know how it turns out.