Principal Machine Learning Engineer

Location: In-person, Mountain View, CACompensation: $250K–$350K base + 1–2% equityActor Labs is one of the first companies taking Vision-Language-Action architectures out of controlled laboratory environments and deploying them onto real-world heavy machinery: excavators, skid steers, and the kinds of machines that actually move the physical world. Closing the gap between a tabletop VLA demo and a policy driving a 40,000-pound excavator through a real worksite is the problem we wake up to every day.The RoleYou'll join a team of under 10 people and work directly with the CTO to own the model side of the company end to end, from data operations through training, evaluation, and deployment back onto the machines. You'll shape how we collect and curate demonstration data, architect and train policies that don't yet exist in the literature, and get them running on real hardware in real environments. If you want to ship policies the world has never seen before and then watch them operate heavy equipment outside in the dirt, this is the role.What You'll Work OnArchitect, train, evaluate, and ship models for behavior cloning and reinforcement learningBuild and maintain data ingestion, labeling, and curation pipelines for high-quality training datasetsDevelop metrics and evaluation frameworks across open-loop, simulation, and real-world settingsWork with simulation, systems, and infrastructure teams to bring ML models into production autonomous systemsDeploy and troubleshoot models in the field, tackling latency, hardware constraints, and integration challengesWhat We're Looking For3+ years applying machine learning with deep learning frameworks (PyTorch, TensorFlow, JAX) to real-world problems (or equivalent technical knowledge)3+ years building/deploying/maintaining ML models in prod. (or equivalent technical knowledge)Fluency in current literature and methods around learned behavior policiesHands-on experience with behavior cloning and/or reinforcement learningExposure to diffusion policies, Vision-Language-Action (VLA) models, or adjacent approachesPublished work at venues such as ICRA, IROS, CoRL, CVPR, ECCV, ICCV, ICML, NeurIPS, or equivalentHow to ApplyEmail your resume and any relevant project/past work to shashi@labs.actor.