Debugging in an Unfamiliar Codebase + Edge System Design

What it tests

Whether a candidate can navigate and fix real code they've never seen before — then design a production-grade system under deployment constraints that mirror actual customer environments.

Format

1. 1Part 1 (40 min): Candidate is given access to a small, unfamiliar codebase with 2-3 real bugs (logic errors, async race conditions, incorrect edge behavior). They diagnose and fix the bugs with the interviewer watching, narrating their process and writing at least one regression test.
2. 2Part 2 (35 min): System design — candidate designs a deployment architecture for a specific real-world scenario (e.g., 'Design a multi-region edge deployment for a Next.js app serving 10M daily users, with per-region cache invalidation'). Interviewer introduces a constraint mid-way (cold start budget drops by 50%, or a region goes offline).

What to look for

• Debugging instinct — do they read the code before guessing, and do they write a test to prove the fix rather than just assume it works?
• Navigation speed — how quickly can they build a mental model of an unfamiliar codebase under light pressure?
• Edge system design fluency — do they understand trade-offs between cold start, cache consistency, and regional failover rather than just naming infrastructure components?
• Constraint response — when a key assumption changes mid-design, do they adapt the architecture cleanly or start over?

Adaptation guide

Use a small snippet of your own codebase (anonymized) with real bugs introduced. The system design prompt should mirror your actual product's deployment reality — if your customers run Kubernetes, design for that; if they run serverless, design for that. The constraint flip mid-design is mandatory: it is the highest-signal moment of the entire exercise.

Full description