Staff Software Engineer, Ai-core (federal)

Okta Okta · Enterprise · San Francisco, CA · SW Eng - Infrastructure-672

Okta is seeking a Staff Software Engineer for their AI-Core team to build a scalable Agentic AI platform. This role involves designing and implementing backend APIs and services for agent identity, machine-to-machine authentication, knowledge base, memory layer, observability, governance, safety, and orchestration. The position requires strong software engineering experience, exposure to AI/ML or agentic applications, and expertise in distributed systems. The role is focused on building the core infrastructure for AI agents within Okta's federal environment.

What you'd actually do

  1. Design and implement backend APIs and services that make up the agentic platform
  2. Build the agent identity and machine-to-machine authentication system, including credential management and delegated access flows
  3. Build the agent knowledge base and memory layer so agents retain context within and across sessions
  4. Build the observability layer for agents: tracing, cost tracking, audit logs, and dashboards that make agent behavior debuggable in production
  5. Build the governance and safety layer: policy enforcement on tool calls, content filtering, PII protection, and human-in-the-loop approval flows

Skills

Required

  • 5+ years of software engineering experience building production backend systems
  • 1+ years of exposure to AI/ML or agentic applications
  • Strong proficiency in one or more backend languages (Python, TypeScript/Node.js, Go, or similar)
  • Hands-on experience designing and operating distributed systems: APIs, microservices, container orchestration, and serverless technologies
  • A security-conscious mindset around credential handling, trust boundaries, and what can go wrong at integration points
  • familiarity with OAuth, OIDC, or other modern auth patterns

Nice to have

  • Exposure to LLM integration
  • RAG pipelines
  • MCP
  • agent orchestration frameworks like LangChain, LangGraph, or the Claude/OpenAI SDKs
  • Experience with policy-as-code authorization (Cedar, OPA)
  • agent identity patterns
  • building developer-facing APIs and SDKs

What the JD emphasized

  • building a scalable Agentic AI platform
  • agents that run real engineering work
  • novel and high ownership
  • problems the industry hasn't solved yet
  • security-conscious mindset
  • Comfort operating in ambiguous, fast-moving environments where the problem definition evolves alongside the solution and the right abstractions are still being invented

Other signals

  • building a scalable Agentic AI platform
  • agents that run real engineering work
  • novel and high ownership
  • problems the industry hasn't solved yet