Harmful Content Detection (LLM/ML)

Problem Overview

Design an end-to-end ML system to detect harmful content. In LLM-specific variants, the focus is on detecting harmful content generated by or passed through a large language model (e.g., toxic outputs, unsafe generations). The interviewer expects coverage of both classical content integrity/moderation concepts (e.g., rule-based filters, classifier models, training data,…

• 5 candidate-reported follow-ups — the exact probes interviewers asked, with the trigger for each
• The rest of the problem statement — full requirements, constraints, and edge cases
• Approach and trade-offs — what passing candidates did, and the mistakes that sink people