Research Engineer – Embodied AI

About RoboTensor

RoboTensor is a Physical AI research and development company building the infrastructure that turns robot foundation models into specialized physical capabilities.

We work across model adaptation, simulation, benchmarking, and evaluation to make general-purpose models capable of performing real tasks on real machines in real operating environments.

Our goal is to push forward how physical intelligence is trained, evaluated, and improved.

About the Role

We are looking for a Research Engineer – Embodied AI to develop new methods for training and adapting foundation models for physical tasks.

You will operate at the intersection of research and engineering: studying promising approaches, forming hypotheses, implementing them quickly, running rigorous experiments, and determining whether they actually produce better policies.

We are looking for someone who cares less about novelty for its own sake and more about research that translates into measurable improvements in physical capability.

What You’ll Do

• Research new approaches to policy learning, model adaptation, and embodied intelligence.
• Develop and evaluate methods for post-training vision-language-action and multimodal foundation models.
• Explore model architectures, training objectives, data mixtures, representations, and learning algorithms.
• Work with imitation learning, reinforcement learning, offline learning, behavior cloning, and related approaches.
• Design and run controlled experiments to understand what actually improves performance.
• Build high-quality research implementations from papers and first principles.
• Develop new approaches for improving generalization, robustness, precision, and long-horizon task performance.
• Investigate how data quality, quantity, diversity, and composition influence learned behavior.
• Analyze policy failures and develop hypotheses about their underlying causes.
• Collaborate with simulation, data, and systems engineers to turn research ideas into complete training experiments.
• Develop evaluation methodologies that expose meaningful capability differences between models.
• Track and evaluate relevant advances in embodied AI, multimodal learning, and foundation models.

What We’re Looking For

• Strong foundations in machine learning, deep learning, optimization, and experimentation.
• Excellent software engineering ability.
• Experience training models with PyTorch, JAX, or similar frameworks.
• Ability to read a research paper, understand the core idea, and implement it independently.
• Strong intuition for experimental design and interpreting ambiguous results.
• Ability to distinguish genuine model improvements from benchmark noise or evaluation artifacts.
• Comfortable working on open-ended problems without a predefined solution.
• Motivated by building systems that work outside controlled research demonstrations.

Nice to Have

• Experience with embodied AI, robot learning, or sequential decision-making.
• Research experience with VLA models, multimodal models, or foundation models.
• Experience with reinforcement learning, imitation learning, or offline RL.
• Experience with generative modeling, world models, or representation learning.
• Experience training large models or running distributed experiments.
• Publications or meaningful open-source work in relevant ML areas.
• Experience moving a research prototype toward real-world deployment.

What Success Looks Like

You turn uncertain research questions into clear experiments.

You can go from an idea or paper to an implementation quickly, establish whether it works, understand why, and use the result to decide what to do next.

Your work ultimately produces policies that learn faster, generalize better, fail less often, and perform better on real physical tasks.