Principal Forward Deployed Engineer

Principal Forward Deployed Engineer, Okta for AI Agents

About Okta for AI Agents

Okta secures access for 20,000 organizations and billions of users. Okta for AI Agents extends that work to the agentic shift. Deploying an AI agent is not like deploying traditional software. You are putting professional work output into production, and it needs deep integration, continuous tuning, and change management. Every agent needs an identity, a scope, an audit trail, and a way to be shut down when it goes wrong. Most enterprises have not built this yet. We are.

We hire builders who see the cracks in enterprise agent identity that everyone else has learned to live with.

The Role

You embed inside four to five of Okta’s most strategic enterprise customers as their dedicated technical partner for agent identity. You sit alongside their identity, platform, and security engineering teams, develop sample and bespoke code, and own the technical outcome from prototype through production.

You are a builder-consultant. You go past architecture diagrams to code, debug, and ship bespoke agent identity solutions inside the customer’s environment. You ship secure agents faster for the customer, and you feed real field insight back to Okta product engineering.

Responsibilities

• Become the customer’s trusted technical voice on agent security. Sit in their standups, design reviews, and incident reviews. Earn a seat on their architecture review board and security council for agent risk decisions.
• Architect and deploy with the customer’s team. Build Okta’s agent security stack into their infrastructure: Cross-App Access (XAA), Fine-Grained Authorization (FGA), MCP Gateway, and agent client registration. Own the identity, delegation, audit, and kill-switch architecture end to end, and coach their engineers on the patterns.
• Engage senior leadership. Brief the CISO, CIO, identity leaders, Chief AI Officer, and principal architects. Translate token-exchange flows into board-level agent risk, and AI governance mandates into architecture.
• Deliver white-glove deployment. Agents in production with full identity coverage, security review passed, governance requirements met, and posture visibility online. The customer points to you as the reason their agent program is real.
• Keep deployments defensible. Align architecture decisions to OWASP Top 10 for Agentic Applications, NIST AI RMF, and MITRE ATLAS, and to HIPAA, FedRAMP, or SOC 2 where the customer is regulated.
• Wire Okta into the customer’s stack. Connect O4AA to their IdP for human-to-agent links, IGA for agent lifecycle, ISPM for posture, SIEM and EDR for behavior coverage, and policy engines for runtime decisions.
• Build evals and observability. Authorization decision latency, scope sprawl across agents, anomalous delegation chains, audit completeness, kill-switch verification, and rogue agent detection.
• Turn field patterns into product. Extract the recurring gaps from their architects and governance leads, and convert them into reusable modules and roadmap fixes that ship for every other customer.
• Be on site. Regular presence at customer locations. Trust and governance alignment happen in the room.

Requirements

• Engineering pedigree. 7+ years shipping production software, still hands-on in the IDE, with on-call experience and operational maturity in systems that authenticate and authorize at high throughput.
• Identity protocols. OAuth 2.0, OIDC, SAML, SCIM, RFC 8693 token exchange, act claims, CIMD and DCR, DPoP.
• Agent security frameworks and platforms. Working knowledge of OWASP Top 10 for Agentic Applications, NIST AI RMF, and MITRE ATLAS. Familiarity with Python, MCP, A2A, ISO/IEC 42001, and the EU AI Act. Comfortable mapping deployments to HIPAA, FedRAMP, and SOC 2.
• Fine-grained authorization. ReBAC and ABAC with policy engines (OPA, Cedar, OpenFGA, or equivalent), and a working understanding of how agents acquire tokens, call APIs, and delegate.
• AI hands-on. Build production integrations with Claude, ChatGPT, Microsoft Copilot, Agentforce, Bedrock, LangChain, CrewAI, the OpenAI Agents SDK, or MCP servers.
• AI-native development. Daily use of Claude Code, Cursor, GitHub Copilot, or equivalent.
• Customer-facing range. At home in a customer standup and a CISO briefing on the same day. You build trust with senior engineering leaders and you stay in the room when their internal politics get sharp.
• High agency, founder’s mindset. A zero-to-one self-starter who owns outcomes end to end.
• Ability to travel, on occasion internationally, up to 35%