Generalist logoGeneralist logo

Research Scientist in Robotics Pretraining

GeneralistSan Francisco Bay Area (San Mateo) or Boston (Somerville)
On-site Full-time

Clicking Apply Now takes you to AutoApply where you can tailor your resume and apply.


Experience Level

Experience

Qualifications

We seek candidates who possess a blend of technical expertise and a passion for advancing robotics through innovative research. A strong background in machine learning, particularly in the development of large-scale models, is essential.

About the job

About the Role

As a Research Scientist focused on Pretraining, you will develop the foundational intelligence layer for robotics. Our mission involves training expansive robot foundation models utilizing vast multimodal datasets that encompass video, proprioception, action traces, language, and beyond. You will lead and execute large-scale training initiatives that imbue our models with groundbreaking general capabilities applicable across various embodiments, tasks, and environments. Your work will involve deeply engaging with all facets of robotic data.

Key Responsibilities:

  • Design and conduct extensive pretraining efforts for robot foundation models, employing transformer and diffusion architectures.

  • Establish model architectures, objectives, and training curricula that leverage multimodal robotic data, including vision, action, state, and language inputs.

  • Create scalable data mixtures and sampling strategies to effectively utilize petabyte-scale datasets.

  • Direct data collection operations and explore new avenues for dataset sourcing.

  • Conduct ablation studies to uncover insights regarding scaling laws, data quality impacts, and architectural trade-offs.

  • Collaborate closely with ML Infrastructure and Systems teams to enhance cluster utilization, throughput, and reliability.

  • Transform raw robotic interaction data into versatile model capabilities.

Ideal Candidate Profile:

  • Extensive experience in training large transformer or diffusion models at scale, particularly in generative tasks such as language, audio, or video modeling.

  • Proven leadership or significant contribution to multi-node, multi-GPU distributed training initiatives.

  • Experience with scaling laws, optimization dynamics, and understanding large-model failure modes.

  • Strong foundation in PyTorch and comfort in debugging across all layers of the computational stack.

  • Appreciation for empirical rigor paired with rapid iteration speed.

  • Enthusiasm for building general-purpose robot intelligence from foundational principles.


About Generalist

At Generalist, we are dedicated to realizing the potential of general-purpose robots. We envision a future where industries and households thrive through innovative collaborations between humans and machines. Our robots are designed to enhance productivity and facilitate the achievement of more ambitious goals.

About Generalist

Generalist is committed to pioneering the future of robotics by creating versatile, general-purpose robots designed to work alongside humans in various environments.

Similar jobs

Browse all companies, explore by city & role, or SEO search pages.

Tailoring 0 resumes

We'll move completed jobs to Ready to Apply automatically.