Member of Technical Staff, Robot Learning

The Mission

Every machine is built and repaired by skilled hands. The world needs more of them. At Darwin, we're building general-purpose robot technicians that assemble and maintain the machines civilization runs on.

We're starting with data centers, where assembly, commissioning, and maintenance require dexterity, precision, and tool use.

Our robots learn these tasks from demonstrations rather than being explicitly programmed for each workflow. We're a small team building the intelligence, hardware, and company to bring these robots into the real world.

Our Values

• Technical Excellence: We are a technology-first company. We value deep technical understanding, rigorous engineering, and exceptional execution.
• Truth-Seeking: We follow evidence, question assumptions, and care more about what’s true than being right.
• Best Idea Wins: Ideas are judged on their merits, not titles, seniority, or who proposed them.
• Low Ego, High Standards: We welcome criticism, challenge each other directly, and hold ourselves to exceptionally high standards. We treat each other with respect.
• High Agency: We take ownership, move quickly, and do whatever is necessary to make progress even under uncertainty.

What You'll Work On

• Robot Learning: Research, design, and train models and policies for learning complex physical tasks from demonstrations and experience.
• Dexterous Manipulation: Develop learning approaches for precise manipulation, assembly, and tool use.
• Learning Architectures: Design novel architectures and algorithms for robot perception, reasoning, and control.
• Data & Training: Design data collection strategies, training pipelines, and methods for learning from real-world experience.
• Evaluation & Deployment: Design experiments, evaluate performance, and improve policies on physical robots.

What You Bring

• Deep understanding of deep learning fundamentals, including neural network architectures, optimization, representation learning, and learning algorithms.
• Strong proficiency in Python, PyTorch, and NumPy, including experience writing efficient, vectorized code.
• Experience with imitation learning and reinforcement learning.
• Hands-on experience designing, implementing, training, debugging, and evaluating deep learning models.
• Experience developing, training, or fine-tuning vision-language-action (VLA) models for robotic manipulation.
• Experience with diffusion models, flow matching, or other generative approaches to action modeling.
• Strong understanding of modern deep learning architectures, including Transformers, attention mechanisms, and generative models.
• Ability to independently design and experiment with novel model architectures, adapt existing methods, and develop new learning approaches.
• Ability to design rigorous experiments, diagnose model failures, and rapidly iterate on architectures, training methods, and data.
• High agency and ownership, with the ability to take ideas from research through implementation, training, and deployment.

Nice to Have

• Experience with world-action models (WAMs) and model-based reinforcement learning.
• Experience with large-scale model training, distributed training, or GPU optimization.
• Experience with real-world robotics, including robot control, perception, teleoperation, or deployment.
• Publications at leading conferences such as CoRL, RSS, ICRA, NeurIPS, ICML, or ICLR.
• Contributions to open-source machine learning or robotics research.

Work Eligibility

Candidates must be authorized to work in the United States. We do not currently sponsor work visas.