Datadog Interview Questions

Datadog's 20 questions read like its product. 9 are coding problems and several are observability plumbing: build span trees from trace spans that arrive out of order, design log queries alongside a buffered writer, implement a buffered file writer that stays correct when several threads call it, implement DeleteTree when the filesystem gives you only a limited set of calls to work with. If you have written ingestion code before, this loop will feel familiar; if you have only ground through LeetCode, the emphasis on buffering and ordering will not. The machine learning questions are implementation-level too, including writing grouped-query attention rather than describing it. Behavioural rounds ask you to explain one project concisely and then go deep on the same project, which is the exchange worth rehearsing out loud. 16 of the 20 came from Software Engineer loops and 15 were technical screens, so this is mostly the first technical conversation. 12 are medium and 5 hard, 8 open in a runnable console, and 11 have a written solution.