Meta Machine Learning Engineer Interview Experience — Four-Round Onsite, Rejected After Two AI Coding Rounds

I applied online to a search/ads/recommendation MLE role. The onsite was four rounds in total: behavioral, system design, and two rounds of AI coding. The behavioral round was just everyday questions, and the system design was a recommendation system. For coding, one question was a triple card game and the other was the maximum number of unique characters, both of them questions you can find online. But I still got rejected because of coding.

On the card game I got to 60% performance, and you have to build the test cases that measure performance yourself, which left me completely blank for a long while...
On the unique characters one, there was one big test case that did not pass.

Lessons I took from it:

1. When you practice, prepare until you can solve it 100%. I thought full marks were not required, so I never even prepared the optimal solution.
2. Use AI more. The recruiter's feedback was "should less manual work, more on AI". I thought I was supposed to implement the simple version by hand first.