We’re building foundational infrastructure to secure AI agents — including their identities, access patterns, and interactions with sensitive systems and data. This includes designing intelligent, dynamic mechanisms for ephemeral access control, secrets management, and agent/user identity tailored to modern agent frameworks such as LangChain, LangGraph, Semantic Kernel, AutoGen, and beyond.
You’ll help define how agents (both machine and human-facing) authenticate, receive scoped access, perform actions securely, and leave behind a verifiable audit trail.
Responsibilities:
- Develop secure, scalable Python services to support agent identity, secrets access, credential management, and authorization flows.
- Implement JWT-based agent/user authentication, and real-time policy checks based on agent context and tool usage.
- Build SDKs, wrappers, and tool integrations that enable popular agent frameworks (LangChain, LangGraph, Semantic Kernel, etc.) to securely request and use secrets.
- Collaborate closely with the architect and other engineers to design components with clear boundaries and clean contracts.
- Ensure secrets and credentials are injected only when needed, redacted from logs, and never persist in agent memory or prompts.
- Write thorough tests and maintain high-quality, well-documented code.
- Work cross-functionally with internal platform, AI, and security teams to understand requirements and refine implementation plans.