I applied through a recruiter. The process took two months.

The whole process is very streamlined and automated as much as possible. Good:

• The recruiter was very nice and generally helpful.

Bad:

• All interviews are fully remote. That's not a problem except for the design interview. It's very difficult to discuss design remotely, especially given hard constraints.
• There are no ML questions, even for an ML engineer role. All technical interviews are only coding. The final design interview was very PM-style, with a strong focus on business metrics. Every time I wanted to discuss ML-related technicalities (i.e., network architecture, loss function, source of training data, etc.), I was harshly interrupted by an interviewer.
• There is zero diversity. All interviewers are young Eastern Asians in their 20s.

Screen:

• One round of 90-minute coding with 4 related problems of increasing complexity.
• LeetCode-style interview with a person with 2 questions: palindrome, trees.

Full loop (each interview is ~45 mins):

• AI-assisted coding where you need to understand a codebase, fix bugs, and implement new functionality. You can use Claude.
• 2 medium LeetCode problems.
• Behavioral: focus on past projects, conflict resolution, etc.
• ML design.