Using stan in an online / production setting

I have plenty of experience getting stan / rstan to do what I need it to in various settings, and it was an easy choice for some form of large scale complex IRT model to handle data at the institutes I’m working at. One of these institutes operates an online learning / testing platform, and wants to transition to using something like the models I’ve already developed, but in an online context where there could be thousands of computations needed in parallel. This is outside the scope of my expertise, but I wonder whether someone out there has been doing anything like this that would be happy to chat on it briefly, for a fee if appropriate?