具身Infra工程师（Seed Robotics） - Seed Model
Core
Building and optimizing large-scale GPU infrastructure and simulation frameworks for training and deploying multimodal large models (Word Action Model, VLA) and reinforcement learning agents in robotics.
Role type
Senior Infrastructure Engineer (Embodied AI / Robotics)
Builds
Distributed training clusters, parallel simulation environments, and edge-side inference engines for robotics.
Domain
Robotics, Embodied AI, High-Performance Computing
Deliverable
production ML models
Required skills
CUDA programming, Python, C++, distributed training frameworks (Megatron, DeepSpeed), reinforcement learning frameworks (RLLib, SampleFactory, Isaac Gym), model deployment tools (TensorRT, TVM, OpenXLA), physics simulation environments (Isaac Sim, MuJoCo, Gazebo)
Preferred skills
Robot hardware deployment experience, open-source embodied AI projects, NVIDIA GPU programming expertise, large model inference and diffusion/flow matching frameworks
Responsibilities
Designing GPU parallel simulation and strategy inference RL training frameworks; optimizing distributed training efficiency for long-sequence Transformers and diffusion policies; developing efficient on-device inference engines for low-latency VLA model execution.
