S
SafetyKit Tech Stack
AI agents for fraud detection and trust & safety workflows
SafetyKit builds AI agents designed to automate fraud investigation, underwriting, and content moderation at scale. The stack is modern and lean—OpenAI, React, TypeScript, Python on AWS—reflecting a team focused on productizing language models rather than building infrastructure. Active projects span classifier systems (dropship detection, IP violation ID) and novel agent applications, while pain points cluster around scaling human review bottlenecks and tuning agent behavior across millions of actions, suggesting the core challenge is moving from prototype to production-grade automation.
Tech Stack 11 technologies
Core StackOpenAI React TypeScript Python AWS AWS Lambda AWS Step Functions Aurora AWS CDK LinkedIn X
What SafetyKit Is Building
◆Challenges
- Fragmented fraud detection systems
- Inaccurate fraud models
- Scaling without toil
- Fragmented monitoring tools
- Custom ml models
- Outsourced review
- Human review bottleneck
- Tuning millions of agent actions
- Operating at scale with limited resources
- Platform stability and performance
▲Active Projects
- Ai agents to investigate signup flows
- Classifier to determine if marketplace products are drop-shipped
- Classifier to identify ip violations deployed on large marketplaces
- Novel ai agent applications
- Risk ai agent workflows
- Model evaluation experiments
- Ai agent software design
- Brand identity establishment
- Customer data interface platform
- Language model productization
Hiring Activity
10 roles · 2 in 30d
Department
Seniority
Company intelligence
