Research, RL Scaling
Core
Researching and scaling reinforcement learning (RL) for frontier AI models, focusing on high-training-compute, long-horizon RL and asynchronous RL to decouple generation from training.
Role type
Senior Research Scientist (RL Scaling)
Builds
Frontier-scale RL training systems and algorithms
Domain
Artificial Intelligence / Machine Learning / Reinforcement Learning
Deliverable
production ML models
Required skills
Python, deep learning frameworks (PyTorch, TensorFlow, JAX), distributed training, asynchronous RL algorithms, inference systems, low-precision numerics, quantization, empirical science, ablation studies
Preferred skills
PhD in CS/ML/Physics/Math, RL for LLMs, distributed training across accelerators, decoupled generation/training systems, RL on agentic tasks, RL training stability, LLM serving stacks (SGLang, vLLM), scaling studies
Technologies
PyTorch, TensorFlow, JAX, SGLang, vLLM, TokenSpeed
Responsibilities
Co-design RL recipes and systems, advance asynchronous RL algorithms, improve rollout generation efficiency, run frontier-scale RL end-to-end, optimize compute and training efficiency, conduct empirical scaling studies
Seniority
Senior, hands-on IC