Machine Learning Engineer — Reinforcement Learning

What you'll do

• Extend distributed training frameworks and integrate training with rollout generation, reward computation, trajectory processing, and policy updates.
• Build and debug multi-node launch, monitoring, logging, checkpointing, and failure-recovery systems.
• Implement and test reliable ML infrastructure and RL pipelines across GPUs and nodes, collaborating with researchers to validate training inputs and outputs.

What they ask for

• 5+ years of experience in ML systems, infrastructure, or distributed training.
• Experience modifying distributed ML frameworks and proven multi-node experience, including debugging skills.
• Strong software engineering fundamentals, including Python, systems design, and testing; experience with an ML platform/infrastructure or distributed inference optimization team.
• Hands-on ownership of an LLM RL pipeline spanning rollout/inference and distributed training integration, plus working knowledge of PPO or GRPO.

Nice to have

• Exposure to mixed-precision training, performance profiling, kernel fusion, or memory optimization.
• CUDA or Triton kernel experience; open-source contributions or published research are also listed as nice-to-haves.

Pay

$150,000 – $450,000/year (base compensation); USD 150,000–450,000 (per-year-salary; base salary)

Tools

DeepSpeed, FSDP, FairScale, Horovod, vLLM, SGLang, PyTorch, JAX

AI focus

Develop and operate reinforcement-learning training infrastructure for foundation models, including rollout generation, reward/verifier integration, trajectory processing, and policy optimization.