interview-system-design

Invoke the skill and pick a mode: batch/warehouse, streaming/real-time, mixed, or ads-domain. An interviewer persona then runs a 60-minute conversational design round with clarifying-question, high-level design, deep-dive, and trade-off phases.

Rounds cover requirements gathering, tradeoff reasoning, and fluency in Netflix systems like Iceberg, Flink, Maestro, Keystone, Psyberg, WAP, Cassandra, and Druid. Interviewers probe failure modes, schema evolution, backfills, observability, security, and 10x scale.

Yes, say "exam mode" at session start to run a pure 60-minute conversational design with no inter-phase feedback. The full scored debrief still runs at the end, which is recommended as a dress rehearsal before the real round.

The interviewer breaks character, delivers a short primer on the concept, and offers to continue or switch to a worked example. Two concept gaps in one session automatically trigger worked-example mode, producing a study artifact instead of continuing the role-play.

The debrief scores seven axes: requirements gathering, tradeoff reasoning, Netflix-tech fluency, failure-mode thinking, schema evolution and backfill, observability and security, and communication. The rubric is calibrated strictly, and missing key behaviors caps the rating at Mixed or Weak.