AI Agent Security Architect: Guardrails & Sandboxing Lead

Replit; Northern, Foster City (KY, CA); Hybrid; USD 180,000 - 240,000; Full time

Benefits offered by this job

401(k) plan with 4% match

Health insurance

Dental insurance

Vision insurance

Life insurance

Short-term & long-term disability

Parental leave

Flexible time off

Commuter benefits

Wellness stipend

In-office setup reimbursement

Quarterly team gatherings

In-office amenities

Job summary

Replit is seeking an AI Agent Security Architect to lead the security blueprint for autonomous agents. You will design runtime defenses, guardrails, and sandboxing architectures that govern agent code execution, tool use, and reasoning across the platform.

You will bridge non-deterministic AI behavior with formal cybersecurity controls for engineering teams and leadership. You will drive risk modeling, threat simulations, and secure design documentation while collaborating with GRC, Sales, and

Qualifications

• 6+ years in security engineering or architecture with 2+ years in AI/ML security or agentic security.
• Deep understanding of agentic architectures and MCP, LangChain, AutoGen, CrewAI, ReAct, and vector databases.
• Hands-on experience with threats unique to agentic AI: indirect prompt injection, hijacking, tampering, data poisoning.
• Strong background in OWASP Top 10 for LLMs & AI Agents, NIST AI RMF, and MITRE ATLAS.
• Proficiency in Python, Go, or TypeScript/JavaScript with security tooling experience.
• Experience documenting and evangelizing security architecture to engineers and executives.
• Excellent ability to communicate complex AI risks to technical and non-technical stakeholders.
• Proven track record contributing to an enterprise Cybersecurity Risk Register.

Responsibilities

• AI Agent Security Strategy & Technical Execution across the platform.
• Define long-term security blueprint for autonomous agent workflows and MCP integrations.
• Design runtime guardrails, semantic firewalls, and real-time verification filters.
• Partner with Infra and AppSec to build secure sandboxed environments for code execution.
• Lead agentic threat modeling and red-teaming for RAG, vector stores, and tool calls.
• architect HITL authorization to enforce least privilege for agents acting for users.
• Implement strict data isolation for vector DBs, RAG pipelines and multi-tenant memories.
• Maintain secure design patterns and SDKs; contribute to risk registers and audits.
• Develop tamper-evident logging and observability for reasoning chains and tools.
• Translate security controls into enterprise audit docs and support sales on inquiries.

Skills

AI/ML security

Threat modeling

Security architecture

Non-deterministic AI risk

Code review & tooling

Security policies & guardrails

GRC & compliance

Communication to execs

Tools

LangChain

AutoGen

CrewAI

ReAct frameworks

Vector databases

WebAssembly

Firecracker

gVisor

Docker

Python

Go

TypeScript/JavaScript

Job description

Replit is seeking an AI Agent Security Architect to lead the security blueprint for autonomous agents. You will design runtime defenses, guardrails, and sandboxing architectures that govern agent code execution, tool use, and reasoning across the platform.

You will bridge non-deterministic AI behavior with formal cybersecurity controls for engineering teams and leadership. You will drive risk modeling, threat simulations, and secure design documentation while collaborating with GRC, Sales, and