Replit
Senior
Build LLM-based guardrails and abuse-detection systems defending Replit's AI-native platform
Replit's Trust & Safety team is hiring a senior engineer to build detection and automated-response systems against phishing, cryptomining, token farming, and prompt injection on an AI-native coding platform. You'll own the full abuse lifecycle from detection to enforcement to appeals, using BigQuery/Hex for investigation and building/fine-tuning LLM-based classifiers. Good prep if you want to practice designing a detection system for an adversarial, constantly-adapting attacker and defending your enforcement-automation tradeoffs.
Step into this interview
Free · a live voice mock calibrated to this exact role
What this interview tests
- Designing abuse/fraud detection systems against an adaptive, adversarial attacker
- Building or fine-tuning ML/LLM-based classifiers for security and abuse detection
- Detecting prompt injection, jailbreaking, and other LLM-specific attack vectors
- Translating investigation findings (BigQuery/Hex) into automated detection rules
- Designing automated enforcement with appropriate human-in-the-loop boundaries
- Owning the full abuse lifecycle: detection, investigation, enforcement, appeals
Common question themes
Design a detection system for a specific abuse pattern (phishing, cryptomining, token farming) on our platform
Tell me about building or fine-tuning an ML/LLM classifier for security or abuse detection
How would you detect prompt injection or jailbreaking attempts against an AI agent
Describe turning an abuse investigation finding into an automated detection rule at scale
Where do you draw the line between automated enforcement and human review, and why
Tell me about owning an abuse or security problem end-to-end, from detection to shipped fix