Senior Researcher - Physical AI & Robot Learning

Fujitsu's RD vision is to create cutting-edge technologies that support society and prioritize the flow of data. Our five key technologies are essential elements for collecting data from all parts of society, transporting it over ultra-high-speed and secure networks, analyzing it with trusted AI, converting it into value, and returning it to society. Central to realizing our strategy and vision are the people behind our RD initiatives. We strive to foster top-notch talent in the innovation process by nurturing skilled individuals equipped to advance this cause with originality and dedication.

Fujitsu Research India (FRIPL) is building a Physical AI Lab to create robots that can perceive, reason, plan, learn, and act reliably in complex real-world environments.

We are seeking an exceptional researcher with a PhD and a strong record of advancing robotics, artificial intelligence, machine learning, or embodied intelligence through rigorous research and physical-robot validation.

The successful candidate will shape novel research in robot learning and control, while connecting perception, task planning, world models, and action execution into dependable end-to-end robotic systems. The role is intended for researchers who can convert ambitious scientific ideas into reproducible, measurable, and deployable capabilities.

• Broad familiarity across reinforcement learning, control, task planning, VLA models, and world models
• Genuine research depth in at least two areas
• Evidence of physical-robot implementation is mandatory

• Research and Algorithm Development
  • Conduct world-class research in Physical AI, embodied intelligence, robot learning, planning, and autonomous decision-making
  • Develop novel methods in:
    • Reinforcement learning
    • Imitation learning
    • Learning-based control
    • Model-based RL
    • Vision-Language-Action (VLA) models
    • World models
    • Long-horizon task planning
  • Design agents that connect semantic reasoning and task planning with motion generation, control, and closed-loop execution
  • Advance robustness, safety, generalization, adaptation, and data efficiency for robots operating under real-world uncertainty
• Physical-Robot Deployment and Evaluation
  • Train, evaluate, and deploy policies on physical robots. Simulation-only experimentation is not sufficient for this role.
  • Address:
    • Sim-to-real transfer
    • System identification
    • Domain randomization
    • Online adaptation
    • Latency
    • Sensing uncertainty
    • Failure recovery
  • Design rigorous experiments, benchmarks, ablations, and evaluation protocols to measure task success, reliability, safety, and generalization
  • Produce reproducible research assets including well-engineered code, experiment documentation, datasets, and robot demonstrations
• End-to-End Physical AI Systems

  • Integrate:

    • Perception
    • State estimation
    • Semantic understanding
    • Task planning
    • Motion planning
    • Learning
    • Control
    • Safety mechanisms

    into complete robotic systems.

  • Collaborate closely with researchers across AI, computer vision, controls, planning, systems, and robotic hardware
  • Contribute to scalable robot-data collection and learning pipelines for manipulation, mobile manipulation, locomotion, or multi-robot settings
• Scientific and Strategic Impact
  • Publish influential research at premier AI, computer vision, and robotics venues
  • Generate patents and contribute to technology transfer, open-source software, datasets, and benchmark development, as appropriate
  • Mentor junior researchers and support research strategy
  • Communicate technical results clearly to expert and business stakeholders
  • Build collaborations with leading academic and industrial research groups

• Completed PhD, or PhD expected before joining, in:
  • Robotics
  • Artificial Intelligence
  • Machine Learning
  • Computer Science
  • Electrical Engineering
  • Mechanical Engineering
  • Or a closely related discipline
• Proven publication record, including meaningful first-author contributions at leading venues such as ICRA, IROS, RSS, CoRL, CVPR, ICCV, ECCV, NeurIPS, ICML, ICLR, IEEE Transactions on Robotics, IEEE Robotics and Automation Letters, Science Robotics, Nature Machine Intelligence, or comparable venues
• Ability to identify important research problems, formulate technically sound solutions, and evaluate them with scientific rigor

• Demonstrated implementation and evaluation on physical robots, supported by publications, project pages, videos, open-source code, datasets, competition results, or deployment outcomes
• Hands-on ability to diagnose the gap between algorithmic performance in simulation and behavior on real hardware
• Simulation-only experience is insufficient

Candidates should have research depth in at least two of the following:

• Robot control
• VLA models
• Foundation models for robotics
• Embodied AI
• Multi-robot systems

In addition:

• Working familiarity with the broader Physical AI stack covering perception, planning, learning, control, system integration, safety, and evaluation
• Strong programming ability in Python and PyTorch or JAX
• Practical experience with C++, ROS/ROS2, and robotics simulation platforms such as MuJoCo, Isaac Sim, or Isaac Lab
• Strong experimental methodology, problem-solving ability, technical writing, and verbal communication

Experience in one or more of the following areas is preferred:

• Robotic manipulation, bimanual or dexterous manipulation, mobile manipulation, locomotion, or human-robot collaboration
• Offline RL, diffusion policies, behavior cloning, self-supervised learning, model-based RL, or test-time adaptation
• Long-horizon task planning, hierarchical policies, memory-augmented agents, neuro-symbolic reasoning, or closed-loop replanning
• Multimodal foundation-model pre-training or post-training, vision-language models, VLA systems, or LLM-guided robotic reasoning
• Sim-to-real transfer, domain randomization, system identification, uncertainty estimation, safety-critical autonomy, or failure recovery
• Large-scale robot datasets, distributed training, scalable evaluation infrastructure, or real-time robotic systems