The process is very structured and well-documented. It typically involves a call with HR, followed by a code pre-screening (a 45-minute LeetCode-style interview), and then four panel interviews. These panel interviews include two LeetCode coding interviews, an ML design interview, and a behavioral interview.

For ML design preparation, the Grokking the Machine Learning Interview course is highly recommended. The behavioral interview is also quite formalized; the guide in "Cracking the Coding Interview" is a good resource.

Unfortunately, I haven't quite reached the bar on the coding interviews specifically. While there's much debate about the relevance of LeetCode-style interviews, I've found them to be especially far removed from the actual experience of an ML engineer. However, these are the rules of the game.

If you are highly motivated to land a job there, practice LeetCode until you lose the will to live, and then a bit more. It has little to do with your normal software development experience as an ML engineer, but this is a special dance you must master if you want to work there.

The organization of the process itself was good, from scheduling to the career portal; I can't complain about those aspects.

For ML design, it was about a recommendation system.

For coding interviews, pick any LeetCode-style medium-level problem and implement it extremely fast. Be sure to correctly assess O-complexity for computation and memory.