NHacker Next
  • new
  • past
  • show
  • ask
  • show
  • jobs
  • submit
Optimizing Recommendation Systems with JDK's Vector API (netflixtechblog.com)
aberoham 1 hours ago [-]
"the remaining 2% were large batch requests", [which made up 50% of the work] .. who really watches that many shows on Netflix? What was in those batches, if someone is watching that much, why bother with serendipity at all? Most serendipitous thing you could do is shut off their subscription.
Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact
Rendered at 20:44:59 GMT+0000 (Coordinated Universal Time) with Vercel.