System design is hard to assess with text alone. A candidate can describe a load balancer in a paragraph, but you learn far more from watching them lay out the pieces — what talks to what, where the data lives, and which calls are synchronous versus fire-and-forget. So we built a place to draw it.
The Architecture Design Studio is a component-graph canvas built into the System Design assessment type. Candidates drag real infrastructure components onto a grid, connect them, and annotate the flow — then write the reasoning behind each decision. The diagram and the written rationale are graded together.
What the candidate works with
The studio is a freeform drawing surface, not a fill-in-the-blanks template. There is no “pick a diagram type” wizard — candidates build the architecture themselves from a palette of components.
- A searchable palette of 29 infrastructure components — clients, web and mobile apps, services and workers, databases and caches, queues and brokers, load balancers, API gateways, CDNs, and more — plus a decision diamond and a text note.
- Sync (solid) and async (dashed) connectors, so the direction and nature of each call is explicit.
- An element inspector to set colour, stroke, and the sync/async style of any connection.
- An Architecture Panel alongside the canvas showing the question, functional requirements, scale and constraints, and the expected deliverables.
The written rationale is required
A diagram on its own is not enough. Every submission includes a written explanation — components, data flow, scaling strategy, and the trade-offs behind each major decision. The rubric grades the diagram and the rationale as one piece of work.
It’s drawing and explaining, not running code
To be clear about what this is: the studio evaluates how a candidate reasons about a system, not whether code compiles. There is no code execution and no sandbox. The candidate diagrams an architecture and defends it in writing — the same thing a strong engineer does on a whiteboard, captured asynchronously so your team can review it on their own time.
How it’s graded
Grading works the way it does everywhere else on the platform. You define the rubric — the criteria that matter for the role. The AI grades the submission criterion by criterion, marks each as met, partial, or not shown, and writes a short explanation with the evidence it drew from. Then a person on your team reviews that reasoning and confirms the result. The AI grades; you decide.
System Design is one of four assessment types live today, all async and all rubric-graded. If you want to see the studio running on one of your own design prompts, request a demo and bring a role and a rubric.