you can't improve the thing significantly anyway plus they are 5 bucks.
Plus if you are just going to copy someones your not making crap, you are copying it. You should know, and have design and board level experience and anyone that has that knows better than to make another 3-5 dollar product that's not tested and has to go through all these approvals to even be able to be sold.
Long story short if you really want to do it:
buy one and trace it. They are 2 layer maybe 4 max.
Well, thanks for the non-answer and all-around discouragement... I guess.
I'm an EE and though my specialty is switchmode power supplies, I know my way around a high speed data bus or two. I am well aware that I could spend hours tracing out the circuit boards but that would be pretty stupid if there are schematics already available. There isn't a lot of uniqueness to the circuit, anyway; it's the implementation that could use some improvement, I feel. More specifically, it seems about 20% of these risers are DOA (lol... China) and a lot of people complain about a loss of hashrate when the board moves from a x16 slot to a x1 riser; I suspect that the USB cable kludge is the culprit so want to try a few other controlled-impedance cable systems out (ethernet interconnects are surprisingly good, for one example, but FFC ribbon cable might be an excellent and very low profile contender).
I am well aware of the technical need for FCC part 15 approval, but if you think any of those risers that you buy for like $8 each actually went through the certification process I got some prime Florida swampland to sell you.