Just finished the Canva senior SWE loop last month and the system design round was the most interesting part. Figured I'd write it up since there wasn't much out there that matched what I actually saw.
First: Canva doesn't use the typical Google/Meta L3-L7 numbering externally. For this post, I'll call it "senior" since that's what the JD said. Internally I believe they map to some band structure but the recruiter wasn't forthcoming.
The format. 60 minutes total with two engineers. One senior, one either staff or a principal equivalent. The first 5-10 minutes are introductions and some quick context-setting. You get the actual design prompt around minute 10.
What they asked me. I can't share the exact prompt but the category was: design a collaborative real-time feature at scale, think Google Docs but for a specific Canva-adjacent context. Heavily focused on operational transforms, conflict resolution, and how you'd handle millions of concurrent editors. This makes sense given their product.
What they cared about. Tradeoffs, not magic. When I said "use Kafka here" they immediately asked why Kafka over SQS or Redis Streams for this specific use case. The answer I gave mattered less than showing I knew WHY each choice had costs.
They also paid attention to how I structured the problem before diving into components. I spent about 8 minutes clarifying requirements and defining the scale before drawing anything. One of the interviewers mentioned afterward (I got debrief feedback through the recruiter) that candidates who rush to the whiteboard often struggle here.
API design. They wanted a clean API contract early. REST vs WebSocket vs gRPC choice was a real discussion point.
Rough comp I was targeting. My offer came in with base around $240k USD for the Austin/remote role. Total package with equity was around $380k annualized but the equity vesting is 4-year standard with a 1-year cliff.
Happy to answer specifics in the thread. Overall thought it was a solid design interview, harder than some mid-tier companies and more practical than some of the more algorithm-heavy shops.