Founding Infrastructure Engineer

Arrows; New York (NY); Sur place; USD 150 000 - 190 000; Plein temps

Résumé du poste

Arrows is building a high-scale agent sandboxing platform on Azure, handling thousands of drawing sheets per project with autoscaling and multi-tenant traffic control.

You will own backend infrastructure, streaming reliability, and the CI/CD lifecycle, while contributing to a Python/FastAPI stack and scalable cloud-native design.

Qualifications

• Strong Python skills for backend tooling and automation.
• Deep production container experience; Azure Container Apps or equivalent.
• Experience with workload isolation and sandboxing of untrusted code.
• Expertise in infrastructure as code (e.g., Bicep, Terraform).
• Ownership of CI/CD pipelines (GitHub Actions or Azure DevOps).
• Azure identity, networking, RBAC, VNets, private endpoints.
• Postgres and Redis operations in production.
• Observability tooling and incident response workflows.
• Experience with AI coding agents and evaluating generated code.

Responsabilités

• Own the agent sandboxing platform and keep it fast, cheap, and secure as agent traffic grows.
• Run the async worker fleet that processes thousands of drawing sheets per project, autoscaling from zero to hundreds of jobs and back.
• Own the LLM traffic layer: keep multi-tenant traffic fair, fast, and metered — every token accounted for in customer billing.
• Keep streaming reliable: chat and progress updates that survive deploys and disconnects.
• Scale document processing: rendering, OCR, and extraction pipelines for large drawings.
• Own CI/CD, reliability, and cost — from the self-hosted runner fleet to incident response to the Azure bill.
• Ship in the codebase: contribute to our Python/FastAPI backend alongside product engineers.
• 4+ years building and operating backend or distributed systems in production, including cloud infrastructure (Azure, AWS, GCP); Azure preferred.

Connaissances

Python

Container experience

Workload isolation

Infrastructure as code

CI/CD

Azure networking

Database ops

Observability

AI coding agents

Outils

OpenTelemetry

Description du poste

You'll own the platform every one of our agents runs on — the execution layer that turns a 3,000-sheet drawing set into hundreds of verified engineering findings. That means the async worker fleet, the agent sandboxes, the LLM traffic layer, and the document-processing pipeline underneath all of it. The headline focus is our agent sandboxing platform: sandboxed code execution for LLM agents— warm pools for instant startup, strict isolation, and controlled network egress for untrusted agent-generated code. We run on Azure Container Apps and Bicep today. We look for deep expertise in containers, workload isolation, and Azure infrastructure rather than any specific toolchain. You will lead our infrastructure decisions going forward.

About the Role

What you'll do: Own the agent sandboxing platform and keep it fast, cheap, and secure as agent traffic grows. Run the async worker fleet that processes thousands of drawing sheets per project, autoscaling from zero to hundreds of jobs and back. Own the LLM traffic layer: keep multi-tenant traffic fair, fast, and metered — every token accounted for in customer billing. Keep streaming reliable: chat and progress updates that survive deploys and disconnects. Scale document processing: rendering, OCR, and extraction pipelines where one project can be tens of gigabytes of drawings. Own CI/CD, reliability, and cost — from the self-hosted runner fleet to incident response to the Azure bill. Ship in the codebase: contribute to our Python/FastAPI backend alongside product engineers.

Responsibilities

• Own the agent sandboxing platform and keep it fast, cheap, and secure as agent traffic grows.
• Run the async worker fleet that processes thousands of drawing sheets per project, autoscaling from zero to hundreds of jobs and back.
• Own the LLM traffic layer: keep multi-tenant traffic fair, fast, and metered — every token accounted for in customer billing.
• Keep streaming reliable: chat and progress updates that survive deploys and disconnects.
• Scale document processing: rendering, OCR, and extraction pipelines where one project can be tens of gigabytes of drawings.
• Own CI/CD, reliability, and cost — from the self-hosted runner fleet to incident response to the Azure bill.
• Ship in the codebase: contribute to our Python/FastAPI backend alongside product engineers.
• 4+ years building and operating backend or distributed systems in production, including cloud infrastructure (Azure, AWS, GCP); Azure preferred.

Required Skills

• Strong Python — our deploy and environment tooling is Python.
• Deep production container experience. Azure Container Apps or similar preferred; AKS depth transfers.
• Workload isolation and sandboxing: Linux primitives (namespaces, seccomp, bubblewrap/gVisor-class tooling) and safe execution of untrusted code.
• Expert infrastructure as code in any major tool — we use Bicep today.
• CI/CD ownership with GitHub Actions or Azure DevOps.
• Azure identity and networking: Entra ID, managed identities, RBAC, VNets, private endpoints.
• Production Postgres and Redis operations.
• Observability: OpenTelemetry, Log Analytics/KQL, incident response.
• Rapid prototyping with AI coding agents (Claude Code or Codex), and the judgment to review and verify what they produce.