All roles
Product Engineering · All three
ML Infrastructure Engineer, Integrations
San Francisco Bay Area · remote-friendly (US)Full-time
About the role
Zero friction deployment is a promise: no recompilation, no hypervisor, no code changes. Keeping that promise across every launcher, scheduler, and framework a customer uses is this role. You will build and test integrations end to end, run pilots with design partners on their clusters, and turn what you learn into product.
What you will do
- Build and maintain integrations for torchrun, Slurm, and Kubernetes launchers
- Validate Watcher and XScope against FSDP, DeepSpeed, and Megatron-LM across versions
- Run design-partner pilots: install, measure overhead, collect failures, report results
- Own the deployment guide and the compatibility matrix
- Feed pilot findings back into the core engineering roadmap
What we are looking for
- Operated large PyTorch training jobs in production
- Hands-on with Kubernetes or Slurm at cluster scale
- Comfortable in Python and shell, able to read C++
- Clear written communication with customer engineering teams
Nice to have
- Experience at an AI lab, GPU cloud, or MLOps platform
- Familiarity with container images and GPU drivers in production