Robotics Machine Learning Engineer

Ixian Robotics•San Francisco, CA

1d•Onsite

About The Position

Ixian Robotics is building intelligent machines to reindustrialize America. The company is based in San Francisco and backed by a16z and Accel. It approaches robotics as a full-stack problem, combining software, hardware, and machine learning with customer understanding, product design, supply-chain navigation, and viable economics. Deployment is central to its innovation process: putting robots to work reveals real-world failures, user needs, and what is required to make machines indispensable. The Robotics Machine Learning Engineer will develop and train models that enable robots to perform physical tasks. The role spans model development, training infrastructure, data curation, and evaluation, with direct responsibility for understanding how model changes affect behavior on hardware. The engineer will design and implement experiments, interpret results, investigate performance issues across models, data, and surrounding systems, and collaborate with software and hardware engineers to determine what to improve next.

Requirements

Experience training deep learning models and explaining the decisions behind experiments.
Strong Python skills and practical proficiency with PyTorch, JAX, or an equivalent framework.
Depth in robot learning or a closely related area, such as computer vision, multimodal learning, or sequential decision-making.
Careful experimental judgment, including selecting useful baselines, making meaningful comparisons, and understanding what limited evaluation data can establish.
Ability to debug training code, inspect datasets, and work directly with physical systems to test conclusions.

Nice To Haves

Experience deploying learned policies on robots.
Experience with imitation learning, reinforcement learning, or vision-language-action models.
Experience with distributed training or building evaluation infrastructure.

Responsibilities

Train and fine-tune models for robotic perception and behavior, selecting approaches suited to the task and available data.
Turn research ideas into working implementations and compare architectures and training methods through controlled experiments.
Curate datasets, investigate data quality, and measure how changes in training data affect performance.
Build evaluations with clear success criteria, repeatable procedures, and regression tracking across model versions.
Run models on physical robots, inspect failures, and work with engineering teams to improve reliability.
Maintain reproducible training workflows, including data loading, experiment tracking, and efficient use of compute.