北京-AI Infra 强化学习工程师（基座研发方向）(J101228)
Core
Develop high-scalability distributed reinforcement learning training frameworks and platforms for large-scale model training and inference optimization.
Role type
Senior IC reinforcement learning infrastructure engineer (distributed systems)
Builds
Distributed RL training frameworks, large-scale post-training platforms, and heterogeneous resource scheduling systems
Domain
AI Infrastructure / Distributed Systems / Reinforcement Learning
Deliverable
production ML models
Required skills
Python, Go, C++, PyTorch, DeepSpeed, Megatron, MSSwift, distributed training, parallel strategies, large-scale system scheduling, reinforcement learning algorithms
Preferred skills
veRL, Slime, vLLM, SGLang, K8s, Ray, Reasoning RL, Agentic RL, open source contributions, top-tier conference papers (OSDI, SOSP, NSDI, MLSys, NeurIPS, ICML, ICLR)
Responsibilities
Develop decoupled RL training frameworks for efficient scaling; optimize asynchronous RL training paradigms and inference scheduling; build fault-tolerant large-scale post-training platforms; design heterogeneous resource scheduling systems; construct observability and automated experimentation platforms
Seniority
Senior, hands-on IC