The fastest-growing job in AI,
all in one place.

Kong — the API gateway company every backend engineer has run into — now runs a full FDE ladder (Forward Deployed Engineer through Senior Staff), embedding with enterprise customers on "API and AI connectivity platform" migrations. 8+ years required across the ladder; this is infra-company FDE, not agent-building FDE — Kubernetes, Terraform, and gateway migrations instead of prompts.

Second new OpenAI vertical FDE req this month: healthcare, requiring real Epic/HL7/FHIR fluency, not just LLM chops — a tell that the FDE motion is now being staffed by domain experts first, engineers second. Same $185K–$300K band as the Legal seat, but with 6+ years required (vs. 5+) and 50% travel.

OpenAI is cloning its FDE org by vertical, not just by geo: this is a dedicated Legal seat alongside brand-new Healthcare and Semiconductor reqs — the same org-chart move Anthropic made two weeks ago, now visible at the other frontier lab too. Published band ($185K–$300K) is real money for embedding with law firms.

The clearest sign yet that OpenAI is building the same FDE org-chart Anthropic revealed two weeks ago: a dedicated people-manager seat over the FDE bench, paying $302K–$385K — more than the $185K–$300K vertical IC seats (Legal, Healthcare) posted the same month. Management layers arriving this fast means the FDE function has outgrown "a few founding hires" at both labs simultaneously.

OpenAI's first standalone Tokyo FDE seat — mirrors the APAC build-out pattern Anthropic already ran (Tokyo) and Sierra runs today (Tokyo infra seat). Requires bilingual Japanese/English fluency, so this isn't a remote-from-anywhere role — it's a bet on a Tokyo-based enterprise practice.

OpenRouter — the de facto default LLM routing layer for a huge chunk of the indie/startup AI stack — posting its first FDE-flavored seat is a tell: infra companies that used to be self-serve-only are now embedding with customers the same way frontier labs do. The job is unusually infra-native: routing, fallback, caching, and cost optimization instead of the usual "build agents" mandate.