Senior Infrastructure Engineer

Superlinked builds SIE, a self-hostable inference engine for pre-trained models under 40B parameters. It packs a wide model catalog (embeddings, rerankers, extractors, OCR) onto shared GPUs at high utilization, so customers serve them in their own cloud for a fraction of managed-API costs. Search and document processing is our initial wedge; the engine is general. Apache 2.0, SOC2 Type 2, 1.8k stars and growing. Headquartered in San Francisco, backed by Index Ventures and other top VCs ($13M seed). The role You will help build SIE and own how it gets deployed and operated. That covers the customer-facing infra that ships with our product (Terraform modules, Helm charts, deployment guides, air-gapped installs) and the internal platform behind it: CI/CD, multi-arch GPU images, model caching, eval infrastructure. Today we deploy to AWS and GCP; Azure and the neoclouds are next. As we move toward managed clusters, you'll set the operational bar that makes a managed offering credible: SLOs, on-call, scale-from-zero behavior, observability, GPU cost. We work in 4-week milestones with high autonomy and an agent-first development process (we use billions of tokens per week). Senior ICs own a requirement end-to-end and multiple projects within it. Who succeeds here You've run large-scale systems in production and know what a real SRE practice looks like: error budgets, runbooks that hold up at 3am, capacity decisions backed by data. You've done this at startups, so you know when shipped beats polished, and you flag slippage early rather than late. You're comfortable working agent-first: delegating real work, reviewing output critically, and recognizing where agents currently fall short. Strong on Kubernetes, Terraform, GPU infrastructure, and one systems level programming language (Rust preferred, alternatively Go / C++ / Zig). Multi-cloud experience helps. Open-source contributor experience a plus. Logistics Hybrid remote. We prefer Hungary, Israel, or the UK for occasional in-person work (1-2x per year), but we're open to the right candidate anywhere in those time zones. This search is for an indefinite-term contracting role. Apply with a short note on the largest production system you've built and operated, what broke, and what you did about it. Bonus points if it ran on GPUs.