There are a few ways to attack this problem, all of which are non-trivial esp. if these people have never done it.
I work for a startup hedge fund and the approach we've taken is parameterizing the vol surfaces, then recompute the prices when we need them. It's important to note that we don't do high-frequency trading, that would likely take an entirely different approach.
The issues that someone would face building out something like that would be dependent on their knowledge of how options should be priced. You also end up having to code up quite a bit of stuff in order to get correct vols (e.g. zero-rate curve, discrete div model). One nice thing is, I believe quantopian would have less headache on the data distribution front, because AFAIK you're able to distribute vol surface parameters (the same is true with yield curves) but not the actual data (stupid I know).