Physical AI Researcher
Software Engineering, Data Science
Montreal, QC, Canada
Posted on Sep 10, 2026
Responsibilities
- As a member of Haply's Physical AI team, you will work directly with researchers and engineers on emerging problems at the intersection of robotics, machine learning, simulation, and haptics.
- Depending on your background and the research problem, your work may include:
- Building and configuring simulated robotic environments for training, data generation, and evaluation.
- Developing end-to-end robot-learning and Physical AI pipelines.
- Working with and fine-tuning foundation models for robotic applications.
- Implementing and evaluating methods from the latest robot-learning and Physical AI research literature.
- Developing and training imitation-learning, reinforcement-learning, diffusion-based, and other learned robotic policies.
- Developing human-in-the-loop learning frameworks that allow operators to demonstrate, guide, intervene in, or correct robotic policies during training and deployment.
- Collecting demonstrations and interaction data through haptic teleoperation.
- Building pipelines for collecting, synchronizing, processing, and training on multimodal robotic datasets, including vision, robot state, actions, force/torque, haptic signals, and human input.
- Investigating how force and haptic information can be incorporated into multimodal foundation models and learned robotic policies.
- Deploying and evaluating learned policies on physical robotic systems.
- Designing experiments to understand and reduce the Sim2Real gap.
- Evaluating and implementing new methods emerging from the rapidly developing Physical AI and embodied AI research ecosystem.
- The role is inherently hands-on. You should be comfortable moving between machine-learning models, Python and C++ code, GPU training pipelines, simulation environments, experimental datasets, and physical robotic systems.
Must Have
- Familiarity with Physical AI and embodied AI workflows, including developing, training, and evaluating robotic systems across simulated and real-world environments.
- Experience with the NVIDIA robotics and simulation ecosystem, including creating and configuring simulated scenes and robotic environments.
- Familiarity with foundation models for robotics, including working with pretrained models, fine-tuning models for downstream tasks, and integrating them into robotic learning workflows.
- Practical experience with modern deep-learning frameworks, particularly PyTorch, including model training, evaluation, and inference.
- Experience working with robotics and machine-learning datasets, including data collection, preprocessing, dataset management, and training pipelines.
- Experience with modern machine-learning experimentation workflows, including tools such as Weights & Biases (W&B) or MLflow for experiment tracking, visualization, model evaluation, and management of training runs.
- Proficiency in Python and working knowledge of C++, particularly for robotics, simulation, or machine-learning applications.
- Experience developing in a Linux environment and familiarity with modern collaborative software-development practices, including Git, debugging, testing, and version control.
- An intermediate background in robotics, including an understanding of fundamental concepts such as kinematics, dynamics, coordinate frames, robotic control, and motion planning.
- Hands-on experience with robotic hardware, including setting up and configuring robotic systems, operating robots safely, interfacing with sensors and actuators, and troubleshooting hardware/software integration.
- Experience working across the simulation-to-hardware workflow, including deploying and evaluating robotic behaviors or learned policies on physical systems.
- The ability to read, understand, and implement methods from contemporary robotics and machine-learning research, and to rapidly evaluate and integrate new approaches as the Physical AI ecosystem evolves.
- A Bachelor's, Master's, or PhD in robotics, computer science, electrical/computer engineering, mechanical engineering, machine learning, or a related field—or equivalent practical experience—is preferred.
Nice to have
- Experience working with multimodal and temporal robotic data, including the collection, synchronization, processing, and representation of signals such as vision, robot state, actions, force/torque, and haptic information.
- Experience with robot and sensor calibration, coordinate transformations, camera calibration, force/torque sensing, encoders, or other robotic sensing systems.
- Experience with real-time robotic systems, including considerations around control-loop frequency, communication latency, synchronization, and deterministic or low-latency software.
- Experience with AWS or other cloud computing platforms, particularly for machine-learning training, data storage, experiment management, or distributed computational workloads.
- Experience deploying or training large machine-learning models using GPU-based computing infrastructure.
- Experience with ROS/ROS 2 or similar robotic middleware.
- Experience with robotic manipulators, teleoperation systems, or contact-rich manipulation.
- Experience collecting and managing robot demonstration datasets, particularly datasets generated through human demonstrations or teleoperation.
- Experience with Sim2Real techniques, such as domain randomization, system identification, dynamics randomization, or adaptation methods.
- Experience developing or working with haptic interfaces, force-feedback systems, or bilateral teleoperation.