Sr. Physical AI Research Scientist

Human Resources•Toronto, ON

9d•CA$140,000 - CA$180,000•Hybrid

About The Position

LG Toronto AI Lab is looking for a Sr. Physical AI Research Scientist who will contribute to designing and developing robot agents that operate safely and robustly in the real world. This role focuses on advancing physical intelligence agents that perceive, reason, and act under real-world constraints, while explicitly addressing safety, uncertainty, and reliability, which are key priorities for LG Electronics as a leading global corporation in the design and development of robotics systems (from industrial and home robots to humanoids). Our mission is to develop safe robotic agents that robustly align their actions with high-level intent, environmental constraints, and safety requirements, even under uncertainty in open-world settings. In this role, you will contribute to: Robot Safety Architecture: Design and implement safety verification and monitoring modules within the robot agent architecture Safety and Generalization: Develop methods to improve the safety, robustness, and generalization of robotic foundation models, including vision-language-action (VLA) models and world models Scaling via Simulation: Build and maintain large-scale simulation frameworks leveraging generative models and domain randomization to enable scalable robot learning Continual Learning: Develop safe exploration and continual learning approaches that enable physical agents to adapt to unstructured environments while respecting strict safety constraints Simulation-to-Reality Transfer: Develop scalable, safety-critical simulation pipelines to rigorously evaluate and stress-test robot agents prior to real-world deployment You will collaborate with researchers and engineers across multiple teams within the organization to build agents that are not only capable, but also safe, reliable, and robust in real-world environments.

Requirements

PhD in Machine Learning, AI, Robotics, or related fields with 3+ years of post-graduate R&D experience, OR M.Sc. in ML/AI/Robotics with 6+ years of experience
Strong publication in top-tier conferences or journals (incl., NeurIPS, ICLR, CVPR, CoRL, ICRA) related to Computer Vision, Continual Learning, Reinforcement Learning, and Robotics
Strong foundation in natural language processing (NLP), computer vision, reinforcement learning (RL), and robotics
Deep expertise in GenAI foundation models, including but not limited to multimodal large language models (MLLMs), vision-language models (VLMs), diffusion and flow matching models
Hands-on experience with constitutional AI, alignment, optimization, and reward modeling/shaping
Proven ability to design, execute, and rigorously evaluate experiments in complex ML/AI systems
Strong proficiency in PyTorch and/or JAX
Highly creative problem-solver with a demonstrated ability to develop novel, real-world solutions
Comfortable operating in fast-paced, technically complex, and highly collaborative environments
Open to feedback from senior colleagues across the organization and actively incorporates input into work
Strong communication skills, with the ability to clearly articulate complex research ideas to both technical and non-technical stakeholders

Nice To Haves

Knowledge of transfer learning, meta learning, and contrastive learning is a plus
Experience with robotic manipulation and navigation tasks in simulation and/or real-world settings
Expertise in robotic foundation models, including diffusion policies, VLAs, and (latent) world models
Pre-training and post-training of robotic foundation models are second nature to you
Familiarity with ROS1 and/or ROS2 for robot software development, system integration, and deployment pipelines is a plus
Practical experience with simulation platforms such as Isaac Sim, Genesis, and MuJoCo
Ability to identify, formulate, and define novel research problems
Experience of collaborating with academic and industry partners
Proven track record of translating research concepts into production-grade prototypes and deployed systems
Strong commitment to mentoring junior researchers and fostering a high-performing R&D culture

Responsibilities

Quickly turn research concepts into practical, validated implementations
Contribute to data collection strategies and training pipelines
Develop imitation learning, reinforcement learning, and/or multimodal learning algorithms
Fine-tune robotic foundational models, including but not limited to VLMs, VLAs and world models
Design safety-aware robotic manipulation solution via including but not limited to fine-tuning, policy steering, and Best-of-N sampling
Develop methods for Sim-to-real transfer, data-efficient learning (offline RL, self-supervised learning), robustness to noise and distribution shift
Collaborate with Physical Intelligence, Embodied AI, and Robotics teams across the organization to develop prototypes and deploy models on physical robots
Support and contribute to R&D collaborations with academic and industry partners
Mentor junior team members and contribute to research direction
Additional duties as assigned