Member of Technical Staff, Microsoft Robotics (Robot Learning)

A member of the Microsoft Robotics team focusing on Robot Learning is sought to develop, train, evaluate, and deploy machine learning models that enable robots to perceive, reason, and act in the physical world. The role spans the full robot learning stack—from data pipeline construction and model architecture experimentation to large-scale training on GPU clusters and real-world deployment on physical robots. The emphasis is on vision-language-action models and related approaches, leveraging imitation learning, reinforcement learning, and other techniques to create reliable, generalizable robot behaviors for manipulation, navigation, and human-robot collaboration.

Microsoft’s mission is to empower every person and every organization to achieve more. We value a growth mindset, collaboration, respect, integrity, and accountability, and strive for a culture of inclusion where everyone can thrive at work and beyond.

Responsibilities

• Develop and train end-to-end robot learning models, including the vision-language-action family, imitation learning policies, and reinforcement learning agents for manipulation, locomotion, and navigation tasks.
• Build, maintain, and optimize data pipelines for robot learning, including telemetry for teleoperation demonstrations, data preprocessing, augmentation, quality filtering, and dataset versioning.
• Train machine learning and deep learning models on GPU clusters, implementing distributed training, hyperparameter optimization, curriculum learning, and automation of training infrastructure.
• Deploy trained models to physical robot platforms, conduct real-world evaluation, debug sim-to-real transfer issues, and iterate on performance based on deployment feedback.
• Implement and maintain evaluation frameworks for robot learning models, including standardized task benchmarks, success rate tracking, generalization testing across objects/environments, and regression detection.
• Collaborate with robotics researchers, simulation engineers, and platform engineers to improve the end-to-end model development lifecycle from data collection to deployment and monitoring.
• Write production-quality Python code (including NumPy, PyTorch, JAX) that is well-tested, maintainable, and extensible, following team coding standards and best practices.
• Review code and technical designs, providing feedback to develop other engineers’ skills and ensure adherence to coding patterns, security practices, and engineering excellence standards.
• Stay current with state-of-the-art research in robot learning, foundation models for robotics, and physical AI, evaluating new model technologies for adoption and integration into the platform.
• Contribute to internal knowledge sharing through technical documentation, brown bag sessions, blog posts, and mentoring of team members.

Qualifications

• Required

• Bachelor’s Degree in Computer Science or a related technical field and 2+ years of technical engineering experience with coding in languages such as C, C++, C#, Java, JavaScript, or Python, or equivalent experience.

• Ability to meet Microsoft, customer, and/or government security screening requirements, including the Microsoft Cloud Background Check upon hire/transfer and every two years thereafter.

• Preferred

• Master’s Degree in Computer Science or a related field and 3+ years of technical engineering experience with coding in languages such as C, C++, C#, Java, JavaScript, or Python, or Bachelor’s with 5+ years of experience, or equivalent.

• Experience in end-to-end robot learning, including imitation learning, reinforcement learning, or vision-language-action model training and deployment on physical robots.

• Proficiency in Python with deep experience in PyTorch, JAX, or TensorFlow for training and deploying deep learning models.

• Experience with robot learning data pipelines, including teleoperation data collection, preprocessing, augmentation, and quality curation for model training.

• Hands-on experience deploying learned policies on physical robot platforms, debugging sim-to-real transfer challenges, and evaluating model performance in real-world settings.

• Familiarity with robotics middleware (ROS/ROS2), robot control interfaces, and sensor processing for perception-action loops.

• Track record of following state-of-the-art research in robot learning, foundation models, and physical AI, including familiarity with leading robotics models and emerging technologies.

• Experience with distributed training on GPU clusters and infrastructure such as Azure Machine Learning or Kubernetes.