1. Phone Screen

Even though it was with a recruiter, we delved into details about my background and past experience to an extent beyond what I had expected. Everything on the resume is fair game. The recruiter will ask lots of follow-up questions and request details.

1. Coding Challenge (I think it was HackerRank)

I'd say this was LeetCode Easy/Medium. I think there were three problems to solve within a one-hour time limit. I could not pass every test case for the second problem, but I moved on to the next round anyway.

1. Coding Interview (With a person who isn't a Palantir employee, using a third-party service)

After scheduling this round of interview, the recruiter sent me an email with a guide on what kind of topics I could expect. The ones I remember are Arrays, Trees, BFS, and DFS. Again, LeetCode Easy/Medium level.

1. Virtual On-Site: Three 30-minute interviews, each with a different task/focus: Coding, Decomp, and something they call "Learning Assessment".

• Learning Assessment: You can't really prepare for this, as it's a test of how you learn new tools/languages/frameworks on the spot. It was really fun.
• Decomp: Practice system design questions. The important thing is to have a "RFC" (Request for Comment), which means by the end of this interview, you want to have a complete proposal such that you can request comments from other people. This is where I kind of screwed up, as there were a lot of ambiguous parts, and you need to quickly decide which part you want to simplify and which part you want to delve into detail. Iterate from a simple but complete solution to a more sophisticated one. I tried to have a complete solution from the start and ended up with an incomplete solution. Have a timer in hand.
• Coding Interview: LeetCode Medium/Hard level. I don't think the questions are reused at all, so you can't prepare for this. In my case, it was an n-ary tree problem.

In summary, it was long and challenging, but I had lots of fun, and everyone that I talked to was really nice. For some reason, for the virtual on-site, I could only choose from the British time zone. Since I'm in Australia, their 9 AM was my 6 PM, so not the best time zone to work with.

An n-ary tree can represent the granting and revoking of access privileges to arbitrary data. For example, you might grant access privileges for North America data to Bob. Later, you may decide to revoke Bob's access for New York data. Then, you might grant access to California data to Amy. How would you implement granting and revoking using an n-ary tree? Be sure to consider the time complexity tradeoffs between granting and revoking.