Some great responses here. I am going to take a slightly different approach. So, machine learning, right? You're trying to build a business that creates value in surfacing correlations in data that are difficult or impossible to do manually (or something to that effect). Great.
Do you have any market validation that it's time to jump straight into building the algorithm? In my (not directly hands on, but repeatedly nearby managing projects like this) understanding, there is quite a lot of work that goes into data acquisition, data cleaning, manually sorting through and guessing different correlations and weights that a data scientist would do that are way before actually implementing any algorithms. It seems like these bits of value that come from the statistics in the data could almost certainly be de-risked with more of a manual or concierge product way before you get to hiring your friend from Google. Have you pitched or acquired any beta customers to try this thing? Can you prove that there is a need for it?
I'll bet you could get a way with an MVP that has almost no actual AI in it. I could be totally wrong, but that's what I have experienced being around data science the last few years.