It builds the Agent Cloud and supplies the infrastructure for AI labs and widely used consumer and enterprise agents, including Manus, Genspark, Lindy, Groq, and Artificial Analysis. Its team includes original authors of Firecracker, the AWS microVM technology behind Lambda, and alumni of Cognition, JetBrains, Zapier, and Wish.
The company has started billions of Firecracker sandboxes across regions on Google Cloud, AWS, and Azure, for customers whose products stop when those sandboxes stop. Here, reliability is solved in code. The fixes for the hardest incidents are code changes, not configuration changes.
This role owns the platform's reliability. You will define what healthy means, build the observability and automation that keep the platform there, and make new capabilities run reliably in production. You will work with the infrastructure and systems engineers who build the platform, and with the sandbox team in Prague.
Responsibilities
Define SLOs and error budgets for core paths such as sandbox start, resume, and connect, and use them in alerting, releases, and prioritization.
Build observability and alerting that shows platform health before customers notice a problem.
Take new platform capabilities from first release to reliable production operation.
Automate operational work, from release checks to routine fleet operations.
Build the security and trust layers the platform depends on, such as service-to-service mutual TLS.
Make private cloud and enterprise deployments as reliable as the company's own regions.
Run incident response, and change how incidents are handled so a class of failure does not recur.
Take on-call shifts and help shape the rotation for the wider platform team.
Take high ownership: write code and design docs for the hardest changes in your area, and ship every day.
Work with SLOs, error budgets, eBPF, Kubernetes, multi-region systems, and noisy neighbor problems.
Requirements
Have run large-scale distributed systems in production across regions, and made their reliability measurable.
Write strong Go. Rust or C/C++ is also welcome. Take ownership of the services you build, and fix issues across the codebase, including in services you did not write.
Have defined SLOs and error budgets for a production service and held the platform to them.
Have been on call, fixed what paged you, and written post-incident reviews that led to lasting fixes.
Have operated Kubernetes in production.
Have used infrastructure as code and GitOps, such as Terraform and Argo CD, with deep experience on at least one of GCP, AWS, or Azure. The company runs on all three, so you do not need all of them.
Have solid Linux and networking fundamentals, enough to debug a host under load and reason about DNS, TCP, and L4/L7 load balancing.
Have operated relational databases and caches in production, including diagnosing performance problems.
Often comes from about five or more years of relevant work, but the depth of what you have built matters more than the number of years.
Bonus: running observability pipelines at scale, such as Prometheus, Grafana, and OpenTelemetry.
Bonus: kernel tracing and eBPF tooling.
Bonus: capacity planning and cost efficiency across clouds.
Bonus: chaos engineering or deterministic simulation testing.
Bonus: familiarity with Firecracker, KVM, or how VM workloads behave on shared hosts.
Bonus: contributions to CNCF or similar open source projects.
Conditions
Work in person in San Francisco or Prague, Czech Republic, with 4 days on-site and 1 day working from home.
Comfortable with the company's code and infrastructure being public.
The company is a fast-growing startup with 9-figure revenue, working directly with top AI companies such as Manus, Genspark, Lindy, Groq, and Artificial Analysis.
The company welcomes applicants regardless of background.