← Verkada Interview InsightsSoftware Engineer
I jumped straight to a relational model with a join table and the interviewer seemed fine with that, but then they pushed on what happens at scale.
Start by clarifying requirements and scale, then propose a data model that supports efficient bidirectional lookups (e.g., a many-to-many mapping between persons and doors). Discuss trade-offs between normalized relational tables and denormalized indexes, and outline how to handle grants, revocations, and queries at scale.
Pro tip: Mention that access control is security-critical, so you should consider caching, eventual consistency, and audit logging; also, proactively discuss how to handle revocation and propagation delays.
Ask about expected number of users, doors, and access checks per second; whether reads or writes dominate; and consistency requirements (e.g., immediate revocation).
Propose a many-to-many relationship between Person and Door, with an AccessGrant entity. Consider using a relational database with a join table, or a wide-column store for scale.
Ensure efficient lookups in both directions: index by person_id and door_id. Discuss denormalization or materialized views if needed for performance.
Describe how to add or remove access, including idempotency, conflict resolution, and propagation to edge devices or caches.
Discuss partitioning, replication, caching strategies, and how to ensure low-latency access checks at scale, including fallback mechanisms.
AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.