Screening stage is an interview with Karat. Standard format as posted in other interview feedback. 5 or so system design questions, and then a multi-part coding question where they will ask time and space complexity. Seems like you need to get the first part of the coding question done and make a start on the second part. I felt unsure about how I did afterwards, but apparently I did pretty well. I think perhaps they're pretty lenient on this one, as I was pretty unsure about at least 1 of the system design questions, and straight up said the wrong complexity for the coding question.
IMO the issues start here. Atlassian's interview material for subsequent stages stresses the importance of clean code, writing tests, and communication. The Karat interview (at least the coding part) is the opposite of this, it's just get the right solution at all costs.
Next stage was two coding interviews. First one was apparently 'Code design', and they give me a variant of the Knapsack problem. Honestly not really sure how this is meant to evaluate 'code design', it's a straight up recursive or DP solution that I couldn't remember in the time allotted. Brute force might have been acceptable, it's really hard to tell.
Second interview was data structures and algorithms. I overcomplicated and also didn't do well at this one. I managed to communicate the right solution after the interviewer helped me realise that I was going down the right path, but didn't get much further than this.
Seemed like there might be the possibility of a down-level to Senior after my interview performance, but that didn't materialize.
Pros: my recruiter was really responsive and kept me informed and in the loop at every stage.
Cons: I think Atlassian need to either be straight up about the fact that they're just asking Leetcode questions, or change their interview practices. My recruiter and some of their prep material IMO de-emphasised things like optimality and sort of said 'well the Karat interview is more Leetcode style, where as the interviews with Atlassian are more 'real world''. The implication was that there was a focus on writing clean code and good unit tests. I don't think this is really accurate. The questions are straight up Leetcode, they just have nice explanations to contextualize them and make them seem like 'real world' problems. Perhaps I just got the wrong impression, but you can't say 'we're more interested in clean code and good tests' and then give me a question straight out of university CS, which would have been trivial if I'd revised it in the last week and is pretty tough otherwise.
To sum up, I think the process if well run, and the questions are ultimately fair questions. I might have been misleveled at Principal, but I also feel like I would have prepped harder if Atlassian just said 'we do Leetcode style interviews', where as I feel like they were trying to portray more of a 'we ask real world type questions'.