This is a Principal Engineer role in the Pathfinding Studio, part of the Engineering Innovation Office, focused on Physical AI system modeling.
As a principal contributor, you will set research direction across foundation model work, vision-language-action models, reinforcement learning and generative models, and lead teams spanning virtual modeling, data curation, learning, perception, planning, control and moving research into hardware.
You will have wide freedom to pick goals and methods, work through hard problems with incomplete information, and guide long-term technology and intellectual property efforts whose results reach several functions and future products.
Responsibilities
Set the long-term Physical AI research direction and system architecture for humanoid and industrial robots, shaping future technology and product plans.
Choose promising new methods and lead work on foundation model training and fine-tuning, VLA models, reinforcement learning and generative modeling, linking perception, planning, control and manipulation into end-to-end robot systems.
Build and maintain virtual world models that cover articulated hands and arms, sensors, robot dynamics, contact and environments.
Own the architecture of Physical AI data pipelines: simulated and synthetic data, real-world and teleoperated collection, multimodal synchronization, schemas, quality checks, dataset versioning, and reproducible training and evaluation.
Lead imitation learning, reinforcement learning, domain randomization, and sim-to-real and real-to-sim workflows; set evaluation criteria, analyze failure modes, and fix complex model and controller problems with limited information.
Turn research into hardware through algorithm-hardware co-design.
Assess system-level performance and resource trade-offs to inform deployment decisions and long-term priorities.
Mentor engineers and researchers across disciplines.
Requirements
8 years of relevant research and engineering experience, together with a Ph.D.
Strong expertise in training and fine-tuning foundation models, VLA models, reinforcement learning, and generative and world models for Physical AI systems such as humanoids and industrial robots.
Strong grounding in mathematics, physics and software engineering, applying rigorous modeling and analysis to complex robotic systems.
Deep knowledge of robotics algorithms and system modeling: kinematics, dynamics, perception, planning, control, grasping, manipulation, trajectory generation, and contact-rich interaction.
Hands-on experience with NVIDIA Isaac Sim, MuJoCo, or similar physics-based simulation and virtual modeling platforms.
Experience building Physical AI data pipelines, including simulated and real-world collection, teleoperation, synthetic data, multimodal data quality, and sim-to-real and real-to-sim workflows.
Prior experience moving research into hardware through algorithm-hardware co-design, validated against system-level requirements.
Experience in digital ASIC design and verification, and in working with research, system architecture, RTL design and verification teams.
Strong software development in Python and C/C++ on CPU and GPU environments, with version control, testing and reproducible experimentation.
Publications in reputable robotics conferences and journals.
Advantage: knowledge of neuromorphic and in-memory computing and modeling.
Advantage: experience with foundation model development and fine-tuning for specification-to-hardware generation, such as schematics and RTL.
Education: a Ph.D. in Robotics, Engineering Science, Computer Science, Electrical Engineering, Computer Engineering, Mechanical Engineering, or a closely related field.
Conditions
The expected annual pay range is $190,000 to $240,000. The final offer depends on geographic location, applicable experience and skillset.
This posting is for a new vacancy.
You must be eligible under U.S. export control regulations, including any export license required before you start, which the company will obtain at its sole discretion.
The company must follow U.S. and Canadian export control laws. Current U.S. law requires an export license before employing people with access to certain technology, and every position here has such access, so candidates are asked about their most recent country of citizenship or permanent residence.
Answering the export control questions is voluntary, but without an answer the company cannot consider your application. Answers are used only to decide whether a license is needed.
Benefits
A competitive benefits package alongside salary, with details shared during the hiring process.
Eligibility for bonus opportunities and commission pay.