DYNA Robotics
Research Engineer
Redwood City, CA · On-site · Full-time · Posted Mar 31, 2026
- Experience
- 7+ years of engineering experience, including leading projects in HPC or ML infrastructure.
About the job
Own end-to-end training and inference infrastructure for embodied AI models, from processing robot data to running models on hardware.
What you'll do
- Scale distributed training across a multi-cloud GPU fleet and build research code and scheduling systems.
- Create high-throughput pipelines for multimodal robot data and optimize GPU utilization, I/O, and memory.
- Deploy low-latency inference for real-time robot control.
What they ask for
- 7+ years of engineering experience, including leading projects in HPC or ML infrastructure.
- Deep knowledge of PyTorch and distributed training, including mixed precision and gradient accumulation.
- Hands-on experience with cloud GPU environments (GCP/AWS) and Kubernetes.
- Understanding of distributed systems, including race conditions, memory management, NCCL, and inter-node communication.
Nice to have
- Robotics data formats (MCAP, Protobuf) or multimodal models such as VLAs.
- Custom kernels (Triton), compilers, or runtime optimization.
Tools
FSDP, ZeRO, Kubernetes, SLURM, TensorRT, Triton, PyTorch, DeepSpeed
AI focus
Build infrastructure to train and deploy embodied AI models for real-time robot control, using multimodal robot data.
Model-written summary · the original posting is authoritative