That's a great question!

You can find more details in section 3.4 in the match cutting paper here:

https://arxiv.org/pdf/2210.05766.pdf

We'll share more about vector search in an upcoming blog post.

If you're interested, also check out the blog post introducing Marken, which is the search backend that we use:

https://netflixtechblog.com/scalable-annotation-service-marken-f5ba9266d428