Senior Researcher - Embodied AI/Robotics - Microsoft Research

Microsoft Research is seeking exceptional researchers passionate about making robots a helpful and safe part of everyday reality. As a member of our team and its collaborations, you will do ground-breaking research at the junction of decision-making (imitation learning, RL, planning) and large multimodal models to enable robots to act and learn robustly in weakly structured environments. 

Responsibilities

• Conduct research in multimodal model training and inference as well as data-efficient learning for decision making to advance state of the art in mobile robotic manipulation. 

• Build proofs of concept that demonstrate the effectiveness of your approaches on physical robot hardware. 

• Actively contribute to team projects as well as collaborations with Microsoft-internal and external partners. 

• Engage in open research activities, including publishing your research findings.  

Qualifications

Required qualifications: 

• PhD in AI/ML, physics, statistics, mechanical engineering, or closely related areas. 

• Well-established track record of research publications, open-source software releases, and other scientific activities. 

• Demonstrated expertise with deep learning frameworks such as PyTorch and strong Python programming skills. 

Other Requirements: 

As a part of you application, please submit the following materials as attachments:

• Your CV. 

• Two conference, journal, or arXiv articles, book chapters, or equivalent writing samples that best demonstrate your expertise. 

• (Optional) The names of two references. 

Preferred qualifications: 

• Experience with at least one project with a publicly available codebase where the candidate has contributed to LLM/VLM/VLA pre- or post-training. 

• At least one top-tier-venue publication with empirical evaluation done by the candidate on physical robot hardware. 

• Experience with at least one project with a publicly available codebase where the candidate has contributed to implementing an imitation learning, RL, or planning approach. 

• Experience with MuJoCo or a PhysX-based simulation toolkit such as Isaac Sim. 

• Demonstrated ability to collaborate, communicate effectively,<