Machine Learning Engineer – RL & Multimodal Post-Training (Physical AI)

Location: Remote within the US or onsite in Columbus, Ohio. The posting also references New York, New York, US, and offers H1B transfers and green card sponsorship.

Company

Venture-backed physical AI company building foundation models for industrial robots, with more than 50 autonomous systems in production and paying industrial customers.

What you will do

• Apply reinforcement-learning post-training methods, including RLHF, GRPO, DPO, and diffusion-based approaches, to VLA and VLM models.
• Train models on image, video, and force/torque data to improve perception and decision-making.
• Design the full model lifecycle, from data collection and fleet learning to continuous learning from production data.
• Build, optimize, and deploy models across a fleet of autonomous systems.
• Collaborate with perception, controls, hardware, and software teams to move research into production.

Requirements

• Industry experience training and deploying reinforcement-learning or vision/multimodal models in production.
• Hands-on experience with RLHF, GRPO, DPO, or comparable post-training pipelines for vision or multimodal models.
• Experience training or fine-tuning VLMs or VLAs at scale.
• Strong Python and PyTorch skills.
• MS or PhD in computer science, robotics, electrical engineering, mechanical engineering, or a related field.
• Work eligibility: remote work is limited to the US; H1B transfers and green card sponsorship are available.

Nice to have

• C++ experience.
• Robotics or simulation experience with Isaac Sim or MuJoCo.
• Publications at venues such as NeurIPS, ICLR, ICML, CoRL, or RLC.
• Experience in big-tech research labs, generative AI companies, or autonomous vehicles.

Culture & Benefits

• NY/SF-level compensation.
• Comprehensive medical, dental, and vision coverage.
• 401(k), flexible PTO, and up to 14 weeks of fully paid parental leave.
• Free lunch for onsite employees.
• Collaborative environment where ideas are welcomed regardless of seniority.