Did two Databricks behavioral rounds across two different interview processes (first one I didn't get past the debrief, second one I got the offer). Let me compare what changed.
First time, I prepped generic STAR stories. You know the type: "I had a conflict, I resolved it, everyone was happy." That did not land well. The Databricks behavioral round felt like they wanted to probe for genuine technical judgment in ambiguous situations, not polished conflict-resolution scripts.
Second time, I leaned into:
Technical influence without authority. They really probe this. "Tell me about a time you got buy-in for a technical decision you believed in but others pushed back on." Not just what the decision was, but how you built the case. What data did you use? How did you handle being wrong about part of it?
Navigating ambiguity at speed. Something like: "Describe a project where the requirements changed significantly mid-execution. How did you adapt?" They want to see you're comfortable making calls without full information.
Failure stories that are actually failures. Not "what I thought was a failure but turned out fine." Actual failures with actual costs. What you learned and what you changed. The interviewer pushed me pretty hard when I gave a softball version. When I gave a real answer about shipping a migration that caused downstream impact for two other teams, they visibly engaged more.
The underlying values they seem to screen for: ownership, directness, building trust through delivery (not just promises), and intellectual honesty. They're not a company that rewards political savvy. They want people who say what they think.
One note: the behavioral interviewer at Databricks was an actual engineer on the team I'd be joining, not an HR screener. That changes the dynamic. They know the work. Generic answers don't fool them.