Member of Technical Staff - Training Platform

Role overview Build a hosted platform that lets developers launch fine-tuning jobs on managed GPU clusters through an API or a simple interface. The work spans the developer-facing product and the Kubernetes infrastructure behind it, supporting both LoRA and full fine-tuning workflows.

Responsibilities - Design and operate training and inference orchestration across multi-cluster, multi-cloud GPU fleets. - Create and maintain Helm charts that package trainers, inference services, environment servers, and related components into reproducible training setups. - Build platform capabilities that make it straightforward for users to start and manage fine-tuning runs. - Improve the infrastructure that executes training jobs on managed GPU clusters. - Work across platform and infrastructure concerns to make training workflows dependable and repeatable.

Requirements - Experience developing platform software, infrastructure systems, or both. - Familiarity with Kubernetes-based workload orchestration. - Ability to build and maintain deployable service configurations, such as Helm charts. - Understanding of machine learning training or inference workloads on GPU infrastructure. - Interest in improving the experience of launching and operating model fine-tuning jobs.

Benefits and work setup - Full-time position based in San Francisco.