At Sakana AI, we are advancing world-class R&D through unique generative AI techniques inspired by nature. Sakana Fugu, our Multi-Agent System as a Model, launched publicly in 2026 and now used by our customers globally, now has a fast-growing userbase. We are looking for an Infrastructure and Platform Reliability Engineer to keep Fugu fast, dependable, and cost-efficient as that moment turns into sustained scale.
You’ll be the person who knows Fugu best, not just how it’s supposed to work, but how it actually behaves under pressure, where it’s fragile, and what it costs to run. Fugu runs on GCP: GKE, Vertex AI, Terraform, all of it. This is a high autonomy, high ownership position, where you will work very closely with both the core R&D and Product Team.
Key Responsibilities
- Language ability: It is preferable if the candidate can communicate in English at a level required to collaborate on technical projects with international team members. Japanese language ability is a huge plus, since the role is based in Japan, but not strictly required. If you are a native Japanese speaker, or passed JLPT N1 or JLPT N2, please do mention this in your application.
- GKE & core infra: the GKE cluster, the gateway service, databases, and every GCP resource behind them, all managed as code through Terraform.
- Model serving: Vertex AI serving infrastructure, GPU allocation and utilization, and keeping serving costs under control as demand keeps climbing.
- Monitoring & alerting: the dashboards and alerts that tell us something’s wrong before customers do. You’ll tune thresholds, kill noisy alerts, and add coverage where we’re flying blind.
- Incident response: triage, diagnose, fix or roll back, and write the postmortem that makes sure it doesn’t happen the same way twice. You’ll be part of the on-call rotation.
- Batch & data pipelines: BigQuery jobs, usage/cost sync, and the DLP pipeline that keeps data handling clean.
- Deploys & spend: safe CI/CD for infra, model, and app repos, plus keeping an eye on spend and right-sizing resources as we grow.
- Customer-facing issues: when something a customer hits traces back to infrastructure, you’re the one who runs it down.
Required Qualifications
- Real GCP operations experience — GKE, networking, monitoring and logging, not just familiarity from a course or a side project.
- You can debug Kubernetes and infrastructure problems under pressure, not just read about how they’re supposed to work.
- Terraform: comfortable both reading existing infra-as-code and writing new modules.
- An on-call mindset: calm when things are on fire, biased to action over analysis-paralysis, and able to write an incident summary someone outside the room can actually understand.
- Enough comfort reading code across app and model repos to trace a symptom back to its root cause, even in parts of the codebase you didn’t write.
Preferred Qualifications
- Experience with LLM or AI-serving infrastructure - Vertex AI, or comparable model-serving stacks.
- Security experience - Cloud Armor, DLP, or similar.
- An SRE background - SLOs, error budgets, runbooks, the whole toolkit.
- Please submit your CV and cover letter in English.
- As mentioned, when applying for this role, please select the Software Engineer (Research and Development, Infrastructure and Platform Reliability) option in Google Forms, and make sure that in your cover letter, explicitly write that you are applying for the “Software Engineer (Research and Development, Infrastructure and Platform Reliability)” role to be considered for this position.