How the interviewer opens it
“Design a service that converts uploaded videos into several formats for online playback. It must process large workloads reliably and keep compute costs within a fixed budget.”
The prompt is vague on purpose. The first minutes are for you. Reduce the prompt to the core features and agree on the numbers before you draw.
What to clarify first
- Scale and source files
- Required outputs
- Playback formats
- Read and write mix
- Latency
- Compute budget
- Consistency and availability
- Durability and retention
- Tenants and invalid media
- Scope limits
The interviewer has the answers. In a SysRound round, Alex gives these answers, as a real interviewer does.
Scope of a strong answer
Core requirements
- Accept uploaded videos and create conversion jobs
- Produce the required video renditions, audio, and manifests
- Report job status and publish complete results
- Recover from worker failures within a compute budget
Usually out of scope
- Live video
- Player implementation and playback delivery
- DRM
- Encoder algorithm development
- Multi-region disaster recovery
What interviewers ask about
- Playback across renditions
- Separate conversions must preserve one timeline. Parallel work can break frame dependencies, segment boundaries, or audio timing.
- Jobs after a worker failure
- Workers can finish media uploads without recording success. Duplicate attempts can continue after a replacement starts.
- Work during a demand peak
- Video duration and codec choice create unequal costs. More workers reduce delay but can violate the compute budget.
Common mistakes
- Treating a queue and a worker pool as a complete design
- Splitting media at arbitrary byte offsets
- Publishing manifests before every referenced object exists
- Repeating all conversions after one task fails
- Increasing worker count without enforcing the compute budget
- Ignoring temporary files, abandoned outputs, and their storage cost
The bar by level
- Senior (L5/E5)
- Defines task dependencies, stale-attempt rejection, and repeatable publication. Explains media boundaries and bounded retries. Enforces tenant shares and both compute limits.
- Staff (L6+/E6+)
- Separates latency promises from budget guarantees. Handles uncertain costs, active reservations, and retry storms. Justifies task granularity and defines recovery after encoder defects.