System Design/openai/Design a CI/CD Execution Platform

Design a CI/CD Execution Platform

HARD45 minDistributed SystemsCi CdJob SchedulingWorkflow OrchestrationMulti TenancySandboxingIdempotencyObservabilityReliability
Reported at: OpenAI

Design a GitHub Actions-like platform with workflow DAGs, isolated workers, artifacts, secrets, retry-safe execution, and regional recovery.

Independent OpenAI candidate reports describe CI/CD and devbox design rounds, including a GitHub Actions-like system where reliability was emphasized before performance. This exercise provides explicit assumptions and scoring around the shared prompt family.

What Candidates Reported

A February 2026 candidate report describes an OpenAI system-design round for a GitHub Actions-like CI/CD platform with high reliability as the first priority. A separate May 2026 Software Engineer report names a devbox and CI/CD pipeline design. The independent accounts support this prompt family without revealing an internal OpenAI rubric.

Problem and Scope

Design event-triggered workflow runs, versioned DAGs, isolated job execution, logs, artifacts, caches, secrets, cancellation, and retry. Clarify job duration, worker types, trust boundaries, acceptable duplicate side effects, retention, and queue-delay objectives before selecting orchestration or compute technology.

Critical Design Decisions

A strong answer separates durable control-plane state from worker execution. Jobs need explicit attempts, leases, heartbeats, and monotonic transitions. Worker pools need fair matching and isolation. Secrets must be short-lived and policy-scoped, while artifacts and caches require provenance, trust-domain boundaries, and retention controls.

Failure Scenarios to Practise

Test worker death after an external side effect, duplicate repository events, cancellation races, a malicious fork, cache poisoning, a huge monorepo burst, exhausted specialized workers, artifact-upload failure, and orchestration-region loss. State which record is authoritative and how uncertain work is reconciled.

How the Answer Is Evaluated

MockMe scores scope and execution semantics (15%), orchestration and workers (25%), security and isolation (20%), scale and reliability (25%), and communication and trade-offs (15%). High scores require durable workflow state, retry-safe attempts, enforceable security boundaries, tenant fairness, and credible regional recovery.

Sources

Ready to practice this question?

Run a mock system design interview with AI coaching and detailed feedback.