Research Engineer - Distributed Training

Role overview Build and improve the distributed systems that support large-scale pre-training and reinforcement learning. Working on a training framework, you will optimize performance across compute, memory, networking, and scheduling, with work ranging from low-level implementation to end-to-end efficiency improvements.

Responsibilities - Develop distributed training infrastructure for pre-training and large-scale reinforcement learning workloads. - Improve training efficiency across compute, memory, networking, and scheduling layers. - Implement low-level optimizations in kernels, communication paths, and runtime components. - Contribute to the training framework that supports these workloads. - Help build the systems foundation used to train frontier-scale models.

Requirements - Strong software engineering ability for distributed training systems. - Experience optimizing performance across multiple layers of a compute stack. - Ability to work on low-level systems such as kernels, communication paths, or runtimes. - Familiarity with pre-training or large-scale reinforcement learning workloads.

Benefits and work setup - Full-time role based in San Francisco or remote. - International candidate assistance, quarterly team off-sites, hackathons, conferences, and learning opportunities.