Urgent hire. Backend-leaning full-stack engineer working directly with the CTO to build and scale core infrastructure for a $30M seed-stage AI platform. 5 days a week on-site at the Presidio, San Francisco. Full-time.

What you'll build

- Core backend infrastructure: agent runtimes, SSE streaming pipelines, task queues, and multi-model LLM provider routing, shipped to real external users.

- Edge compute layer using Cloudflare Workers and Durable Objects with Fly.io blue-green and canary deployments.

- Real-time systems at scale: WebSockets, SSE, event streaming, and message queues under production load.

- Distributed system architecture: consistency tradeoffs, failure modes, and operational runbooks for a 15-person team moving fast.

- On-call rotation: you carry a pager, respond to incidents, and own resolution end to end.

Required

- 5+ years of production backend engineering, including on-call. Can name a specific incident, what broke, and how it was resolved.

- TypeScript fluent, backend-primary. Full stack across the production stack: TypeScript, Cloudflare Workers, Durable Objects, Hono, Drizzle ORM, Postgres, Zod, SST, Vercel AI SDK, Docker, React, Vite.

- Has shipped real-time systems to external users in production: WebSockets, SSE, event streaming, or message queues, with specific scaling bottlenecks they can describe.

- Has designed distributed systems and made concrete consistency tradeoffs. Can explain what they chose, what they gave up, and why.

- Has shipped infrastructure that real external users depended on, not internal tooling. Can name the system, the user load, and the operational responsibility they held.

- Has worked with LLM agent loops, model routing layers, or tool-use architectures. Does not need ML depth, but must not be starting from zero on agent interaction patterns.

- Based in San Francisco Bay Area or willing to relocate before the start date for 5-days-a-week on-site work at the Presidio.

Visa and relocation:

- US work authorization required. O-1 or J-1 sponsorship considered for exceptional candidates only. H-1B is not on the table.

- Europeans eligible for O-1 are considered. US-based candidates are prioritized.

Disqualifiers

- No US work authorization and not O-1 or J-1 eligible.

- Unwilling to work on-site 5 days a week in San Francisco. Not negotiable.

- Pure frontend background with no substantial production backend or distributed systems experience.

- No exposure to agent loops or model interaction patterns. The ramp from zero is too large for this hire.

- Built only internal tooling, never shipped infrastructure to external users.

- Needs written specs and fully async communication to function. The office is small, open, and high-energy. Information flows through direct, impromptu conversation.

Strong plus

- Production experience with Cloudflare Workers, Durable Objects, or serverless edge compute.

- Has built agent runtimes, model routing layers, or tool-use architectures with LLMs in production.

- Ex-founder or ex-CTO. Most of the current team are former founders or CTOs. Signals high agency and comfort with ambiguity.

- Early engineer at an AI-native startup. The existing team came almost entirely from small AI-native companies.

- Actively building or experimenting with AI outside of work, not just following the space. Has tried the product before interviewing.

- Open to a paid in-person work trial of a few days up to two weeks. Roughly 70% of trial candidates convert to full-time. Accommodation and flights covered for relocating candidates.