VLA Full-stack ResearcherVLA 全栈研究员

Mission岗位使命

Close the loop from language and vision to robust action on a real humanoid, across data, modeling, evaluation and deployment.打通从语言和视觉到真实人形机器人稳健动作的全链路，覆盖数据、建模、评测和部署。

Responsibilities工作职责

• Build datasets and representations for multimodal action learning.构建多模态动作学习的数据集与表征。
• Train, ablate and evaluate VLM/VLA policies and action experts.训练、消融和评测 VLM/VLA 策略与动作专家。
• Integrate policies with simulators and real-robot control loops.将策略接入仿真器和真机控制闭环。

Qualifications任职要求

• Hands-on experience with PyTorch and multimodal or sequence models.有 PyTorch 及多模态或序列模型实战经验。
• Understanding of pre-training, post-training, RL or imitation learning.理解预训练、后训练、强化学习或模仿学习。
• Enjoy debugging the full path from tensors to robot behavior.喜欢调试从张量到机器人行为的完整链路。

How to apply申请方式

Send a CV, projects or publications to hr@xvirobotics.com.请将简历、项目或论文发送至 hr@xvirobotics.com。