Job description

FieldAI is transforming how robots interact with the real world. Our growing R&D team is based in Boston, where we develop risk-aware, reliable, field-ready AI systems that tackle the hardest problems in robotics and unlock the potential of embodied intelligence. We take a pragmatic approach that goes beyond off-the-shelf, purely data-driven methods or transformer-only architectures, combining cutting-edge research with real-world deployment. Our solutions are already deployed globally, and we continuously improve model performance through rapid iteration driven by real field use. We are looking for a Research Scientist to advance the state of the art in large-scale learned humanoid manipulation. In this role, you will develop new methods for learned and physically-grounded models, spanning reinforcement-learning, imitation learning, multimodal representation learning, cross-embodiment transfer, and beyond. Working within FieldAI’s broader humanoid manipulation roadmap, you will formulate new approaches, design rigorous experiments, and validate resulting capabilities on real humanoid robots. You will work closely with research engineers and systems teams to ensure that research advances translate into scalable, reliable manipulation and loco-manipulation systems. You will also contribute to FieldAI’s technical direction in robotics foundation models, including how models are architected, trained, evaluated, adapted, and deployed across robotic platforms. This role combines fundamental research with the opportunity to demonstrate meaningful advances on complex real-world robotic systems. What You’ll Get To Do Advance Humanoid Manipulation Research Help lead high-impact research projects in general-purpose humanoid manipulation and loco-manipulation. Develop novel model architectures, learning objectives, action representations, and training methods for large-scale robot learning. Establish strong baselines, evaluation protocols, and benchmarks for manipulation performance and generalization. Advance Robotics Foundation Models Research large-scale VLAs and other multimodal behavior models that connect perception, language, reasoning, and continuous robot action. Develop imitation-learning and reinforcement-learning methods for improving robustness, precision, dexterity, and long-horizon task performance. Investigate model adaptation, temporal abstraction, memory, uncertainty, data efficiency, and compositional skill learning. Study how model performance scales with architecture, data quantity, data quality, task diversity, and embodiment diversity. Develop methods for transferring capabilities across robots while accounting for embodiment-specific sensing and control constraints. Drive Real-World Deployment and Validation Lead research projects from initial hypothesis through large-scale training, deployment, and validation on physical humanoid robots. Design real-world experiments that expose model limitations and measure generalization under realistic environmental variation. Analyze failures at the level of data, representations, policy behavior, perception, and control. Use insights from robot deployments to guide new research questions and model improvements. Demonstrate learned capabilities across dexterous manipulation, bimanual coordination, and loco-manipulation tasks. Translate Research Into Scalable Systems Partner with research engineers to turn new algorithms into reproducible training pipelines and reliable robot capabilities. Help define data-collection strategies, dataset composition, evaluation standards, and model-development priorities. Provide technical leadership across research projects and mentor other researchers and engineers. Communicate findings clearly through internal technical reviews, publications, presentations, and open-source releases where appropriate. Balance longer-term research efforts with advances that support FieldAI’s near-term autonomy roadmap. What You Have PhD in Robotics, Computer Science, Machine Learning, Electrical Engineering, Mechanical Engineering, or a closely related field. A strong research record in robot learning, robotic manipulation, reinforcement learning, imitation learning, multimodal learning, or a related area. Demonstrated ability to lead research projects from an initial technical question through rigorous experimental validation. Deep expertise in at least one relevant area, such as foundation models, VLAs, generative policies, reinforcement learning, imitation learning, or multimodal foundation models. Experience training and evaluating modern deep-learning models using PyTorch. Strong understanding of robotic manipulation, including relevant aspects of kinematics, dynamics, control, perception, and planning. Experience deploying and evaluating learning-based methods on physical robotic systems. Strong experimental judgment, including the ability to design informative ablations, baselines, metrics, and evaluation protocols. A track record of publications or equivalent research impact in leading robotics or machine-learning venues. Ability to communicate research ideas clearly and collaborate effectively across research, engineering, systems, and product teams. The Extras That Set You Apart Research experience with humanoid robots, bimanual systems, or multi-fingered robotic hands. Experience developing robotic foundation models, multimodal transformers, diffusion or flow-based policies, or other large behavior models. Experience with reinforcement-learning fine-tuning, offline RL, reward modeling, preference learning, or autonomous policy improvement. Research on cross-embodiment transfer, multi-robot learning, or embodiment-general action representations. Experience developing or studying large-scale robotics datasets and data-scaling strategies. Experience combining autonomous rollouts, simulation, web-scale data, teleoperated data, and other heterogeneous data sources. Experience leading collaborations spanning algorithm development, infrastructure, and real-robot deployment. Why Join FieldAI? FieldAI is tackling one of robotics’ hardest problems: deploying robots in unstructured, previously unknown environments. Our Field Foundational Models™ advance perception, planning, localization, and manipulation with an emphasis on explainability and safety, so our systems can be trusted where it matters most. You will work alongside a world-class team that values creativity, resilience, and bold thinking. We bring a decade-long track record of real-world deployments, strong performance in DARPA challenges, and experience from organizations such as DeepMind, NASA JPL, Boston Dynamics, NVIDIA, Amazon, Tesla Autopilot, Cruise, Zoox, Toyota Research Institute, and SpaceX. Our R&D organization is growing and anchored in Boston, with close collaboration across our teams in Southern California and with colleagues around the US and globally. Be Part of the Next Robotics Revolution Solving problems at this scale takes a team as unique as the mission. We are looking for people who push beyond conventional approaches, enjoy tackling tough and ambiguous questions, and bring interdisciplinary perspective. Our success depends on exceptional AI researchers and engineers, as well as strong software developers, product designers, field deployment experts, and communicators who can turn breakthroughs into real capability. We are headquartered in Irvine, Southern California, with teammates across the US and around the world. Join us to shape the future of embodied intelligence as part of a fun, close-knit team building systems that work in the real world. Equal Opportunity FieldAI celebrates diversity and is committed to creating an inclusive environment for all employees. Candidates and employees are evaluated based on merit, qualifications, and performance. We do not discriminate on the basis of race, color, religion, sex, gender, national origin, ethnicity, veteran status, disability status, age, sexual orientation, gender identity, marital status, or any other legally protected status.

Requirements

Must-haves

What You’ll Get To Do
Advance Humanoid Manipulation Research
Help lead high-impact research projects in general-purpose humanoid manipulation and loco-manipulation.
Develop novel model architectures, learning objectives, action representations, and training methods for large-scale robot learning.
Establish strong baselines, evaluation protocols, and benchmarks for manipulation performance and generalization.
Advance Robotics Foundation Models
Research large-scale VLAs and other multimodal behavior models that connect perception, language, reasoning, and continuous robot action.
Develop imitation-learning and reinforcement-learning methods for improving robustness, precision, dexterity, and long-horizon task performance.
Investigate model adaptation, temporal abstraction, memory, uncertainty, data efficiency, and compositional skill learning.
Study how model performance scales with architecture, data quantity, data quality, task diversity, and embodiment diversity.
Develop methods for transferring capabilities across robots while accounting for embodiment-specific sensing and control constraints.
Drive Real-World Deployment and Validation
Lead research projects from initial hypothesis through large-scale training, deployment, and validation on physical humanoid robots.
Design real-world experiments that expose model limitations and measure generalization under realistic environmental variation.
Analyze failures at the level of data, representations, policy behavior, perception, and control.
Use insights from robot deployments to guide new research questions and model improvements.
Demonstrate learned capabilities across dexterous manipulation, bimanual coordination, and loco-manipulation tasks.
Translate Research Into Scalable Systems
Partner with research engineers to turn new algorithms into reproducible training pipelines and reliable robot capabilities.
Help define data-collection strategies, dataset composition, evaluation standards, and model-development priorities.
Provide technical leadership across research projects and mentor other researchers and engineers.
Communicate findings clearly through internal technical reviews, publications, presentations, and open-source releases where appropriate.
Balance longer-term research efforts with advances that support FieldAI’s near-term autonomy roadmap.
What You Have

Nice-to-haves

None flagged.

Application prep

Use the role’s language and application details while tailoring your materials.

Keywords to mirror

aimachine learning

Platform: lever
Effort: quick