Hybrid Large Behavior Models Research Scientist en Boston Dynamics

As a Large Behavior Models Research Scientist, you will join a world-class team of engineers and scientists focused on building generalized manipulation skills on Atlas. Training generative models on large-scale demonstration data is showing incredible promise for building multitask manipulation capabilities on robots. We are training models that leverage multiple data sources, including teleoperation, human demonstrations, and internet data, to train policies capable of few-shot generalization in tasks that range from highly dexterous to high-force whole-body manipulation. You will play an active role in researching, architecting, training, and deploying learned models to expand the manipulation capabilities of Atlas.

How you will make an impact:

• Creatively apply machine learning to solve real-world bimanual manipulation problems

• Write high-quality (documented, tested, maintainable) Python and C++ code

• Lead design reviews and collaborate closely with members of the Atlas software team

• Deploy and debug your code on Atlas

We are looking for:

• MS in Computer Science, Machine Learning, Robotics, or a related field

• Prior experience training and deploying learned policies for complex behaviors in robots or simulated characters

• Experience with common ML tools, architectures, and workflows

• Strong analytical and debug skills

Nice to have:

• PhD in Computer Science, Machine Learning, Robotics, or a related field

• Experience with modern behavior cloning technology, e.g., diffusion policies, ACT

• Experience working with and contributing to large codebases

• Experience working hands on with robot hardware

#LI-JM1