Coding

Agent Rating System

Asked at

Atlassian

1Give yourself 5 minutes

2Answer out loud, not in your head

3Then compare with the answer below

Stuck? Show a way to structure it+

01Ask what an agent is and who supplies ratings.
02Clarify scale, volume, freshness, and abuse constraints.
03Separate raw events from an aggregate score.
04Define ranking and tie-breaking semantics.
05Discuss monitoring and feedback loops.

Reference answer

The canonical record contains only "Agent Rating System" with no users, events, scale, or desired behavior, so do not invent a particular algorithm. Begin by clarifying whether agents are humans, AI workers, support representatives, or marketplace providers; whether ratings are explicit stars, outcomes, or both; and whether the score drives display, routing, compensation, or safety. A sound baseline stores immutable rating events with rater, target, timestamp, context, and moderation state, then maintains an aggregate such as count, mean, and a confidence-aware score. For public ranking, discuss Bayesian shrinkage or a lower confidence bound so an agent with one five-star review does not outrank a proven agent. Specify idempotency, one-rating-per-eligible interaction, abuse detection, and recomputation. Big-O depends on query shape: event ingestion can be O(1) indexed write, while a precomputed leaderboard lookup can be O(log n) or O(1) by rank store.

Then expect these follow-ups

How would you prevent review fraud?
Tests: follow-up reasoning
When should aggregates be recomputed rather than updated online?
Tests: follow-up reasoning

Hiring now

Get the Atlassian interview, not just the prep

landed applies to roles like these for you, on each company’s own site, with answers written for the role.

Get 5x interviews10 free auto-applies. No card.

Or practice this question first

Next question