Did the Databricks onsite last month for a senior backend infra role. Writing this down because I couldn't find a recent thread on what the final round actually looks like day-of.
Mine was virtual, 5 rounds across one day. Back to back with a lunch break in the middle. Here's the breakdown:
Round 1: Coding (90 min). Two problems, one medium one hard. I wrote about the coding format in a separate thread but in short: communication matters as much as correctness here. They watch how you reason.
Round 2: System Design (60 min). Infra-flavored, as expected. Mine was something in the vein of "design a distributed log/event store" which gave me strong Kafka-meets-Delta-Lake vibes. Leaned into fault tolerance, replication factors, recovery paths.
Round 3: Behavioral (45 min). Real engineer, not HR. Heavy on ownership and navigating ambiguity. (I wrote a separate thread on behavioral if you want the detail.)
Round 4: Domain deep dive (60 min). This was the most Databricks-specific round. They asked me to walk through a past project in detail and then probed the architectural decisions. What would you do differently? If the scale increased 10x, what breaks first? Where are the latency cliffs? It felt like a collaborative design review, not a trap. I liked this round the most.
Round 5: Bar raiser equivalent (45 min). A senior person from outside the team. Broader questions: how do you think about platform-level reliability? Describe your mental model for debugging production issues. Less about specific knowledge, more about thinking style.
Debrief timeline: I got the recruiter call with a verbal offer 5 days after the final round. They told me the debrief takes 2-3 days internally. That matched.
Overall, the hardest loop I've done in terms of sustained intellectual demand. But also one of the better-run processes. Interviewers were clearly prepped on my background and asked relevant questions.