HackTheRounds Interview Experiences

How I Passed the Canva SDE Interview (2026): Take-Home, Coding & System Design

Complete Canva SDE interview walkthrough: HR screen, coding round, take home assignment, and 4 intense final rounds including ML debugging and real time collabo

By Anonymous · 2026-02-20

Background

I applied for a Software Engineer position at Canva. The entire process took about two months from first contact to offer. Canva's interview is unique — they include a take-home assignment that's evaluated on production-quality code, not just correctness.

Round 1: HR Screening (30 min)

Standard intro call focused on communication skills and resume walkthrough. Questions about my project experience, why Canva, and some basic technical concepts. The recruiter was friendly and set clear expectations for the rest of the process.

Round 2: First Coding Round (45-60 min)

A single technical interviewer with a couple of problems:

  • Rate Limiter Design — implement a configurable rate limiter with different strategies (sliding window, token bucket)
  • Longest Subarray with Sum <= K — classic sliding window problem

The interviewer also asked follow-up scaling questions: how would this rate limiter work in a distributed system? What if you need to handle 1M requests/second?

Round 3: Take-Home Assignment

This is where Canva is different. The core challenge was a sliding window / prefix sum problem, but the evaluation criteria were heavily weighted toward code quality:

  • Production-quality code with proper edge case handling
  • Readability — clear naming, modular structure, comments where needed
  • Testing — unit tests covering happy paths and edge cases
  • Documentation

Important: I've heard of algorithmically correct solutions getting rejected because the code wasn't "production-ready." Don't treat this like a LeetCode submission — treat it like a PR you'd open at work.

Round 4: Final Onsite (4 intense rounds)

Round 4a: ML/Debugging Given logistic regression code with poor online performance. Had to identify issues: data distribution mismatch between training and production, overfitting on training set, missing L2 regularization. This round tested debugging intuition more than ML theory.

Round 4b: Engineering Scenario Build an image processing workflow: normalization pipeline, failure handling, retry logic, and an LRU cache for processed images. Very practical — felt like something you'd actually build at Canva.

Round 4c: System Design Design a real-time collaboration feature for simultaneous multi-user design editing (think Google Docs but for design). Key discussion points: - Conflict resolution strategies (OT vs CRDT) - Data consistency across users - Latency optimization for real-time updates - Handling offline users who reconnect

Round 4d: Behavioral Focused on disagreement resolution and teamwork. Canva emphasizes "Be a good human" as a core value — they're genuinely evaluating whether you'd be a pleasant colleague, not just a strong engineer.

Tips

  1. Take-home quality is everything — spend extra time on readability, testing, and documentation
  2. Canva loves practical engineering — expect real-world scenarios, not abstract puzzles
  3. System design = collaboration features — study real-time editing, CRDTs, operational transforms
  4. Be genuinely kind in behavioral rounds — they take "be a good human" seriously