About The TeamThe team works on research and systems that advance frontier models. Our work often goes beyond standard training recipes, which means we also build the infrastructure needed to make new training approaches practical at scale. This is a team where systems work is directly tied to research progress: better tools, abstractions, and runtimes can unlock experiments that would otherwise be too slow, brittle, or difficult to express.
About The RoleThis is a systems engineering role focused on ML training infrastructure. You will work on the systems layer that turns novel research ideas into runnable, measurable training workloads for large models. The work can sit on the critical path for model releases, bringing both the excitement of direct impact and the responsibility of building systems that remain reliable under real pressure.
In This Role, You Will- Build and maintain infrastructure for large-scale model training and experimentation.
- Design APIs and interfaces that make complex training workflows easier to express and harder to misuse.
- Improve reliability, debuggability, and performance across training and data pipelines.
- Debug issues spanning Python, PyTorch, distributed systems, GPUs, networking, and storage.
- Write tests, benchmarks, and diagnostics that catch meaningful regressions.
You Might Thrive In This Role If You- You want to build systems that enable new model training approaches, not just optimize established ones.
- You have strong systems instincts and care deeply about performance, reliability, and clean abstractions.
- You have good taste in API and interface design, with empathy for the researchers and engineers using your tools.
- You are comfortable working across ML research code and production-quality infrastructure.
- You enjoy debugging from evidence: profiles, traces, logs, tests, and minimal reproductions.
