AI Agent Security Architect

Replit•Foster City, CA

6d•$230 - $350•Hybrid

About The Position

Replit is seeking an AI Agent Security Architect to serve as the primary technical authority for the company's autonomous and AI agent security blueprint. This role involves designing, implementing, and maintaining runtime defense systems, guardrail frameworks, and sandboxing architectures for AI agents that execute code, invoke tools, and reason on the platform. The architect will lead high-impact AI security initiatives and communicate complex AI security concepts to both engineering and executive leadership.

Requirements

6+ years of experience in security engineering or security architecture
At least 2+ years dedicated specifically to AI/ML security, LLM application security, or securing agentic frameworks.
Deep understanding of agentic architectures and protocols (e.g., Model Context. Protocol (MCP), LangChain, AutoGen, CrewAI, ReAct frameworks, and vector database technologies).
Hands-on expertise with threat vectors unique to agentic AI: Indirect Prompt Injection, Goal Hijacking, Tool Parameter Tampering, Data Poisoning, and Context Contamination.
Strong background in applying safety standards and risk frameworks such as OWASP Top 10 for LLMs & AI Agents, NIST AI RMF, and MITRE ATLAS.
Proficiency in Python, Go, or TypeScript/JavaScript with a proven track record of reviewing code and building functional security tooling or guardrails.
Experience authoring, maintaining, and evangelizing technical security architecture documentation.
Exceptional ability to communicate complex non-deterministic AI risks to both deep technical engineers and executive stakeholders.
Proven track record of contributing to an enterprise Cybersecurity Risk Register.

Nice To Haves

Experience designing runtime sandboxing and isolation technologies (e.g., Firecracker, gVisor, Docker, WebAssembly) for dynamic code execution environments.

Responsibilities

Define the long-term vision and architectural patterns for securing autonomous agent workflows, Model Context Protocol (MCP) integrations, multi-turn reasoning loops, and multi-agent coordination.
Architect and deploy dynamic input/output guardrail systems, semantic firewalls, and real-time intent verification filters to prevent goal hijacking, system prompt leaks, and indirect prompt injections.
Partner with Infrastructure and AppSec teams to design secure, short-lived, micro-isolated environments (e.g., microVMs, WebAssembly, container sandboxes) for agent code execution and interaction with host operating systems.
Conduct specialized threat modeling against non-deterministic systems and lead automated and manual AI red-teaming initiatives to uncover vulnerabilities in RAG context pipelines, vector stores, and tool-calling interfaces.
Architect fine-grained permission models and step-up Human-in-the-Loop (HITL) authorization patterns for agents acting on behalf of users, ensuring delegated privileges are not exceeded and API keys are not misused.
Design strict data isolation and poisoning prevention controls for vector databases, retrieval-augmented generation (RAG) pipelines, and long-term conversation memories across multi-tenant workloads.
Define, document, and maintain authoritative secure design patterns and SDKs for engineering teams building AI agent capabilities.
Identify, quantify, and document non-deterministic and agentic security risks (e.g., OWASP Top 10 for LLMs & AI Agents, MITRE ATLAS), ensuring accurate representation in the Cybersecurity Risk Register.
Design tamper-evident logging and telemetry standards for agent reasoning chains, tool execution history, and context assembly to support incident response, forensics, and GRC requirements.
Translate complex AI security controls into enterprise-ready audit documentation (e.g., ISO 42001, NIST AI RMF, EU AI Act readiness) and support Sales on enterprise security inquiries regarding AI safety.