腾讯游戏-大模型强化学习框架研发工程师/专家
Core
Design, develop, and optimize training frameworks combining video generation, world models, and reinforcement learning.
Role type
Senior IC machine-learning engineer (reinforcement learning & distributed training)
Builds
Training frameworks for video generation and world models
Domain
AIGC, Reinforcement Learning, Distributed Systems
Deliverable
production ML models
Required skills
Python, distributed training (Megatron-LM, DeepSpeed, FSDP), reinforcement learning algorithms (PPO, GRPO, DPO), RL frameworks (Verl, ROLL, AReal), system programming, complex system debugging
Preferred skills
AIGC project experience, technical curiosity, independent problem solving
Technologies
Python, Megatron-LM, DeepSpeed, FSDP, PPO, GRPO, DPO, Verl, ROLL, AReal
Responsibilities
Design and develop training frameworks integrating video generation and world models with reinforcement learning; Optimize the full RL training pipeline including data loading, memory management, communication efficiency, and parallelization; Collaborate with algorithm teams to validate cutting-edge technical prototypes.
