I applied through a recruiter. I interviewed at Encord (London, England) in Dec 2025
Interview
The first stage is a chat that isn’t very technical, mostly behavioural questions with a technical angle, such as describing projects you’ve worked on.
Next, there is a take home assessment. It takes longer than they suggest. The task isn’t extremely hard but is also not straightforward.
After that you’re invited to a live coding round based on the take home assignment. They advise it is important to build as much as possible as quickly as possible but it is not exactly true. During the interview they insist on algorithmic approach instead. I found this to be a waste of time as they claim to value people who can work fast, but then reject simple and effective solutions in favour of textbook algorithms. This ends up filtering out strong practice driven candidates in favour of people who have the time to grind leetcode.
Given take home task. I misunderstood the brief - went for speed of implementation rather than optimization. Task was interesting and I was given quality feedback afterwards despite not passing the round
I applied through other source. I interviewed at Encord (London, England) in Apr 2026
Interview
A take home test. My main issue with the company is that I've spent time completing the take home test, and they came back to me with a run-of-the-mill rejection letter.
Professional team, but the assessment process feels outdated for an AI role
Pros:
- Respectful Process: The team was professional and organized throughout.
- Good Take-Home: The take-home project was fair, well-scoped, and representative of actual work quality.
- Feedback: The team offered a feedback call after the rejection (rare and appreciated).
Cons:
- Artificial Constraints: The live coding environment prohibits modern tools (Claude, Cursor, ChatGPT, documentation). For an AI-focused company, banning the tools that define modern engineering feels like a missed opportunity to evaluate real world workflow.
- Signal Mismatch: There is a disconnect between the take-home and the live rounds. Candidates invest time in a substantial, practical take-home project, yet the hiring decision seems to filter primarily on algorithm recall rather than the engineering skills demonstrated in the work sample.
- Outdated Patterns: The live coding optimizes for memorizing specific algorithm patterns rather than the problem-solving or iterative testing used in actual product development.
- Industry Comparison: Competitors like Labelbox or Roboflow evaluate via pair programming on real deployment tasks. This process feels closer to generic Big Tech gatekeeping than a modern AI startup assessment.
Advice to Management:
- Weight the take-home project more heavily. It is a much better indicator of the candidate's actual output than the algorithm rounds.
- Update the live coding to reflect modern engineering constraints (i.e., allow AI tools/docs) to see how candidates actually ship code.
Advice to Candidates:
- Prepare for the Algo Filter: The live coding acts as a hard gatekeeper, even a great take-home won't save you if you fail here. Drill classic LeetCode Medium patterns (especially Matrix/Graph problems) and prioritize this over polishing your project.
- Simulate the Constraints: You will be coding without your modern stack (Cursor, Claude, etc.). Practice solving problems in a raw environment without AI or auto-complete to avoid being slowed down by the artificial restrictions.
Interview questions [1]
Question 1
Standard LeetCode style matrix manipulation algorithm (e.g. in-place rotation).
