Staff ML Engineer, Robot Foundation Models (Defense Manufacturing)

An early-stage robotics company building autonomous machines for American defense manufacturers is hiring staff-level ML engineers to train vision-language-action models. The work is applied rather than pure research, with new systems built from scratch and the chance to put models onto real robots nearly immediately.

Scope

• Train and refine VLA models to meet reliability and speed goals for a live customer deployment
• Try different data, training recipes and hyperparameters to discover what lifts task success
• Build and grow distributed training infrastructure on multi-GPU clusters so experiments move quickly
• Own the model lifecycle from assembling datasets through training, evaluation and validation on real robots
• Expand experimentation capacity across several workflows as operations grow

Qualifications

• Six or more years as a machine learning or robotics engineer, with production-ready ML shipped to physical hardware
• Hands-on PyTorch or JAX for training deep learning models
• Experience running distributed training on multi-GPU clusters
• Familiarity with transformer-based, multimodal or generative models; VLA, VLM or robotics foundation model work is a plus, as is imitation learning or diffusion policies
• Ability to own ambiguous projects and deliver working systems; this is not a fit for pure researchers
• A degree in computer science, ML, robotics, EE or a related field, or equivalent experience
• U.S. citizenship, required for defense work

On-site in New York, NY, with no remote or hybrid option. Unable to consider candidates who need visa sponsorship or transfer.