Machine Learning Research Scientist- Atlas Perception

Do you have a passion for applying state-of-the-art deep learning techniques to solve real problems in robotics? As a Research Scientist on the Atlas Perception team, you will play a key role solving the uniquely challenging perception problems in humanoid robotics, including scene understanding, localization, object pose estimation, and terrain prediction. You will work closely with software, mechanical, and electrical engineers to research, develop, and test new robot technologies for Atlas and other R&D efforts at Boston Dynamics.

As a leader in the field you will shape novel ML perception techniques to drastically improve the utility of mobile robots by enabling them to effectively interact with the world around them.

How You Will Fit In

• Focus on demonstrations: make the robot do exciting things that impress the world
• Engineering scrappiness: take research prototypes all the way to robot hardware, often involving unglamorous work
• Goal orientation: focus your research on delivering value to the robot platform
  
  

Day To Day Activities

• Identify high-impact areas where deep learning can be applied to solve real problems for our robots
• Research and evaluate new ML solutions to solve novel vision problems
• Train and adapt existing machine learning architectures for visual scene understanding
• Regularly present research within the Atlas team and in wider company meetings
• Guide and participate in ML and robot infrastructure software engineering projects
  
  

We Are Looking For

• Strong ML / Computer Vision research background
• Experience applying ML to robotics, hands-on robot experience
• Proficiency in Python programming and popular ML frameworks
  
  

Additional Desired Experience

• Experience with ML research in fields such as:
• Model-free 6DoF object pose or grasp estimation
• 3D depth estimation, scene reconstruction, and learned visual representations
• Interfacing perception with learned control policies (RL, Behavior Cloning)
• Object detection and scene understanding w/ VLMs
• Foundation Models for embodied AI
• Experience training and fine-tuning large models with in-house data
• Practical knowledge of 3D geometry, projective geometry, camera models
• Familiarity with typical robotics sensors and calibration procedures (e.g. ToF, Stereo, RGB, IMUs)
• Competent with C++ programming