AI Agent Security Architect
Replit is an agentic AI software-creation platform for creating, launching, and growing apps, websites, tools, and businesses through natural language.
Funding history
About Replit
Replit combines an AI agent with integrated application infrastructure—including authentication, databases, hosting, monitoring, integrations, and enterprise controls—to let users build and deploy software from one workspace.
Skills
Candidate Availability
Required and preferred rules are kept separate and reflect the wording in the original posting.
About the Role
You will define security architecture for autonomous agent workflows, deploy runtime guardrails and policy enforcement, and design sandboxed execution environments. You will lead AI threat modeling and red teaming, implement identity and data-isolation controls, establish secure design patterns and telemetry standards, maintain risk documentation, and support compliance and enterprise security inquiries.
Requirements
- 6+ years of security engineering or security architecture experience
- 2+ years of AI or ML security, LLM application security, or agentic-framework security experience
- Knowledge of agentic architectures and protocols
- Knowledge of Model Context Protocol, LangChain, AutoGen, CrewAI, ReAct, and vector databases
- Knowledge of prompt injection, goal hijacking, tool parameter tampering, data poisoning, and context contamination
- Knowledge of OWASP Top 10 for LLMs and AI Agents, NIST AI RMF, and MITRE ATLAS
- Proficiency in Python, Go, or TypeScript or JavaScript
- Experience building security tooling or guardrails
- Experience authoring security architecture documentation
- Experience contributing to a cybersecurity risk register
Responsibilities
- Define security architecture for autonomous agent workflows and integrations
- Architect and deploy runtime guardrails, semantic firewalls, and intent-verification filters
- Design sandboxed environments for agent code execution and tool use
- Conduct agentic threat modeling and lead AI red-teaming initiatives
- Architect least-privilege permission models and human-in-the-loop authorization patterns
- Design data-isolation and poisoning-prevention controls for RAG and memory systems
- Maintain secure design patterns and SDKs
- Document AI-security risks in the cybersecurity risk register
- Design tamper-evident logging and telemetry standards
- Support compliance documentation and enterprise security inquiries
Benefits
- Equity
- 401(k) program with 4% match for US employees
- Health insurance
- Dental insurance
- Vision insurance
- Life insurance
- Short-term disability
- Long-term disability
- Paid parental leave
- Paid medical leave
- Paid caregiver leave
- Flexible time off
- Holidays
- Commuter benefits for in-office US employees
- Monthly wellness stipend
- In-office setup reimbursement
- Quarterly team gatherings
- In-office amenities