Design Jira bug-to-team classification system
Company: Atlassian
Role: Machine Learning Engineer
Category: ML System Design
Difficulty: medium
Interview Round: Onsite
## Problem
Design a system that automatically classifies incoming Jira bug tickets into the most appropriate owning team, and produces a report for customer companies.
Assume:
- Each ticket includes fields like title, description, component/service, logs/stack traces, priority, and customer/company.
- Historically, tickets were assigned to teams by humans; that history can be used as labels.
- The system should suggest or auto-route tickets to teams, while allowing overrides.
- The system should support reporting (e.g., per customer: bug counts by team/category over time).
## Deliverables
1. Requirements and success metrics.
2. Data sources and labeling strategy.
3. Model approach (baseline to advanced) and feature design.
4. Online serving architecture (routing + human-in-the-loop).
5. Training pipeline, evaluation, and monitoring for drift.
6. Reporting layer for customer-facing analytics.
Quick Answer: This question evaluates skills in machine learning system design, including supervised classification, feature extraction from textual and log data, labeling strategy from historical assignments, human-in-the-loop routing, and customer-facing reporting.