As an Autonomy Engineer focused on VLA (Vision-Language-Action) pre-training, you will train capable robotic policies using deep learning, including pre-training base models on multi-embodiment trajectory data, fine-tuning for specific tasks, and optimizing for real-time edge inference. You'll own the full loop from data collection and curation through deployment, partnering with data and MLOps teams to build continuous training pipelines that ingest synthetic data and teleop logs.… The role requires 3+ years building production deep-learning systems with hands-on experience in LLMs, VLMs, or generative models, plus strong Python and PyTorch/JAX skills to profile, debug, and write maintainable research code.