okay i went through the Renaissance Technologies coding interview process a few months ago and the online assessment specifically caught me off guard. documenting this because the format is not standard.
the OA itself: timed. 3-4 hours depending on how you count. you get multiple problems, the difficulty ramps. what's different from typical leetcode-style OAs: problems are original. i cross-referenced afterwards and could not find any of them on leetcode or codeforces. one section was straight algorithm/DS: trees, graphs, dynamic programming. this part felt like hard-ish leetcode. another section leaned heavily mathematical: probability, combinatorics, some number theory. this is the part that trips people up if they haven't revisited that content since college. code quality seemed to matter. one problem asked for a solution and then asked about edge cases explicitly. felt like they read what you wrote, not just whether it passed test cases.
coding rounds (on-site): also algorithmic. i had two coding rounds on-site. one was more pure DS/algo, one blended in a practical systems-adjacent problem (how do you process a stream of data efficiently). both involved follow-up questions after i got something working. the follow-ups were the hard part. "what if the input size is 10x" or "what if this is distributed" -- they don't always want you to rewrite everything, they want you to reason.
language: they're flexible. python is fine. if you want to impress, C++ helps because they care about low-level performance. i used python and got through fine.
overall difficulty: harder than a typical FAANG coding round. not harder in a "trick question" way, harder in a "real depth" way. if you're LC hard-ready you'll do okay on the algo sections. the math component needs dedicated prep.
the interviewers were good, not mean about it. but the bar is high and there's no faking depth here.