No need to reinvent the wheel. These discussions took place on how Bisq can implement it, how to verify transactions etc. So I would start by looking how Bisq implemented it, then we can think if there are improvements that can be made, e.g. working without a collateral.
It should indeed be clear from the beginning that this would be an ‘idealistic’ project for Seria, likely with little return of investment on the short term.I do not know why we even have low liquidity on Bisq, but if we cannot get high liquidity there, there is no reason to assume it would be much better on Serai.
On a first glance, the systems are similar, with two parties interacting plus a mediator/validator in Bisq who checks if a payment came through. Both Bisq and Seria appear to use a multisig, although Bisq uses a 2 out of 2 multisig, and Serai apears to use a treshold multisig. So there are still some differences to consider. Bisq uses the grinnode live transaction proof verification API, Serai could probably do the same to have an easy way to verify if grin transactions actually came through.
I think Bisq uses a collateral in the 2-2 multisig to protect against spam and to completely cover for cost in case of fraudulent behavior (which was one of the questions by the Serai devs). Normally the coin itself would be used as collateral, and spam protection, but with Grin the coin itself cannot be locked in the 2-2 multisig, so instead they use a BTC collateral, to achieve the same objective.
When you think about it, this is all completely valid and acceptable for a DEX, so again no clear explanation why Bisq has low liquidity nor any reason to assume Serai would get a better liquidity from the community.