System Design
Watch how a candidate reasons about scale and failure. They diagram the system in the Architecture Design Studio and explain it — both graded against your rubric.
Architecture Design Studio
System Design Assessment
Tools
View
Components
29 kindsElement properties
Architecture Panel
Question
Design a URL shortener
Fast and available at high read volume.
Functional requirements
- Create short links
- Redirect to origin
Scale & constraints
Reads
100M / day
Latency
< 50 ms p99
Expected deliverables
- Component diagram
- Written rationale
Design rationale — written explanation graded with your diagram
RequiredA written explanation is required alongside the diagram. Cover the components, data flow, scaling strategy, and the trade-offs behind each decision — the rubric grades the diagram and the rationale together.
The Architecture Design Studio — draw the system, then explain it. Sample content shown.
How this assessment works
What it assesses
How a candidate reasons about scale, data, and failure — the judgement behind a design, not just whether they can name a pattern.
What the candidate does
Reads the design prompt and writes an open-ended response in their own words, asynchronously. Optional prompts can steer the answer toward the dimensions you care about.
How it's AI-graded
The AI reads the written answer and grades it criterion by criterion against the rubric you defined — marking each as met, partial, or missing, with the evidence it drew from. You review the reasoning and confirm the score.
Assessment report · AI-graded result
AI grade
AI generated0 / 10086% confidence
Solid separation of read/write paths; cost trade-offs left implicit.
Scoring rubric
Why this score
The answer cleanly splits read and write paths and justifies the key-generation choice with a scaling argument. Failure behaviour is addressed with a sensible degrade-writes-before-reads stance. Cost and operational trade-offs are named but not weighed.
The result, with its reasoning
The AI grades the answer against your rubric and shows its work: a score out of 100, a confidence read, which criteria were met, partial, or not shown, and a short explanation of why. Every part of this is model-assisted and clearly labelled — the hiring call stays with your team.
How grading works
You define the rubric
Set the criteria that count for this role. The rubric is frozen for the run so every candidate is graded the same way.
The AI grades against it
Each answer is graded criterion by criterion, with the evidence the model drew from and a confidence score — never a black-box number.
You confirm the decision
The score is decision support, not the decision. A person reviews the reasoning and confirms it before it counts.
Built to be trusted
Audit trail
Every grade, override, and confirmation is recorded immutably — who, what, and when.
Your data, walled off
Your rubrics, questions, and candidate data stay scoped to your workspace.
GDPR-ready
Soft-delete by default and candidate anonymization when you need it.
Related reading
See System Design running on your roles.
We're onboarding our first hiring teams. Bring a role and a rubric, and we'll show you the whole flow — candidate answer to confirmed decision.
Request a demo