Member of the Technical Staff - Customer Tools

Andromeda Cluster•San Francisco, CA

1d•Remote

About The Position

Andromeda is hiring a Member of the Technical Staff to build customer-facing software that helps AI labs operate compute across multiple providers as one fleet. The product will give customers a unified way to run jobs and see what every GPU is doing, while allowing them to offer idle GPU hours on Andromeda’s marketplace and reclaim that capacity when needed. This is a new product, and the role will build its first versions in collaboration with principal engineers and customer research and infrastructure teams. The work is primarily backend development in Go, using Kubernetes and Postgres, with a TypeScript console. The backend manages customer clusters. The position includes carrying an on-call pager and speaking with customers every week.

Requirements

Impressive technical work with demonstrable depth and real-world impact; the posting says experience may range from three to twenty years.
Experience building multi-tenant systems, including authorization, organizations, and tenant data separation.
Fluency with Kubernetes, including building against the Kubernetes API rather than only deploying to Kubernetes.
Experience building metrics, logs, or event pipelines that support dashboards and actionable alerts.
Experience shipping production services in Go, or a systems background sufficient to ship production Go within the first month.
Ability to build front-end features for the features you own.
Production on-call experience for a system customers depend on, including making it quieter afterward.

Nice To Haves

Experience with ML or HPC infrastructure, such as training clusters, Slurm, or job schedulers, from an operator’s or researcher’s perspective.
Experience with cloud consoles or developer platforms used by engineers running real workloads.
Experience with enterprise identity, including SSO, SCIM, and identity-provider integrations.
Experience syncing or replicating large datasets across clusters or regions.

Responsibilities

Build a unified interface for customers to run jobs across clusters from Andromeda, other providers, and their own sites.
Develop visibility into cluster health, usage, spending, and alerts.
Build organization, role, and single sign-on capabilities that work across customer clusters.
Enable customers to identify idle GPUs, offer unused capacity on the marketplace, and reclaim it when needed.
Identify customer problems, determine what to build, and ship product features.
Build backend services primarily in Go against Kubernetes and Postgres, as well as front-end features in TypeScript.
Manage customer clusters through the backend.
Participate in production on-call and communicate with customers weekly.