<div><span>Help us build CloudWalks ambitious R&amp;D program around humanoid robotics. Youll own the full cycle of developing locomotion and high-level behaviors: design tasks and environments in NVIDIA IsaacLab/Sim/Gym, train policies with RL/IL, validate on simulation and deploy on our physical humanoid robot (Sim2Real).</span></div><h3>The Robotics Division</h3><ul><li><b>We work on embodied AI and autonomous systems</b> across various robotics platforms and applications</li><li><b>Full development cycle</b> from simulation environment design in NVIDIA IsaacLab/Sim/Gym to deployment on physical robots</li><li><b>São Paulo lab with advanced robotics hardware</b> including humanoids and vending machines</li></ul><h3>What youll do:</h3><ul><li><b>Invent and build humanoid tasks and environments</b> – from simple flat-ground locomotion to challenging obstacle courses</li><li><b>Train and refine policies</b> using RSL-RL, RL-Games, Stable-Baselines3, or similar GPU-parallel RL libraries</li><li><b>Explore cutting-edge approaches</b> – reinforcement learning, imitation learning, diffusion policies, and transformer-based approaches for continuous control</li><li><b>Close the Sim2Real gap</b> through domain/dynamics randomization, calibration, and staged deployment on the humanoid platform in our lab</li><li><b>Integrate with robotics ecosystem</b> – ROS 2 (ros2_control, MoveIt 2) and NVIDIAs robotics stack (Isaac ROS, cuRobo/cuMotion, NVBlox)</li><li><b>Orchestrate complex behaviors</b> with <a href="http://BehaviorTree.CPP/Groot">BehaviorTree.CPP/Groot</a>, and experiment with Isaac GR00T for high-level humanoid skill composition</li><li><b>Measure and iterate</b> with reproducible pipelines, dashboards, and design documentation to accelerate R&amp;D cycles</li></ul><h3>What youll need:</h3><ul><li><b>Solid ML and RL foundations</b> (PPO, SAC, TD3) with hands-on PyTorch experience</li><li><b>NVIDIA Isaac ecosystem proficiency</b> – Isaac Sim/Isaac Lab and legacy environments (Isaac Gym, OmniIsaacGymEnvs, Orbit)</li><li><b>Strong ROS 2 skills</b> – ros2_control, lifecycle nodes, planning and perception integration</li><li><b>Sim2Real expertise</b> – domain randomization, dynamics tuning, and safe hardware deployment</li><li><b>Engineering best practices</b> – version control, containerization, reproducibility, and documentation</li><li><b>Must be able to work on-site in São Paulo</b> for regular robotics lab sessions with physical robots</li></ul><h3>Bonus points if you have:</h3><ul><li><b>Legged locomotion experience</b> with humanoid control tasks (Legged Gym, ETH Zurich RSL environments)</li><li><b>MuJoCo-based RL workflows</b> (Gymnasium, dm_control, robosuite) for algorithm prototyping and benchmarking</li><li><b>Advanced policy architectures</b> – Diffusion/Transformer policies for robotics applications</li><li><b>NVIDIA robotics stack</b> – GPU-accelerated planning (cuRobo, cuMotion) and Isaac ROS perception (NVBlox, VSLAM)</li><li><b>Real-time systems and mechatronics</b> knowledge (PREEMPT_RT, C++ optimization)</li><li><b>Strong portfolio</b> of robotics and ML projects (GitHub repos, demonstrations, research papers)</li></ul><h3>Work Model</h3><ul><li><b>Hybrid role</b> – collaborate remotely but spend regular days in our robotics lab</li><li><b>Hands-on experimentation</b> with physical humanoid robots is a core part of this job</li><li><b>Test, validate, and deploy</b> your simulation work on real hardware in our São Paulo lab</li></ul><h3>Recruiting Process</h3><ul><li><b>Challenge:</b> Youll receive a technical challenge to demonstrate your skills in robotics, simulation, and machine learning</li><li><b>Technical interview</b></li><li><b>Cultural interview</b></li><div><b>Note:</b> If you are not willing to take an online quiz and demonstrate your technical capabilities with both simulation and physical robotics systems, do not apply.</div></ul><div><b>Diversity and inclusion:</b></div><div><span>We believe in social inclusion, respect, and appreciation of all people. We promote a welcoming work environment, where each CloudWalker can be authentic, regardless of gender, ethnicity, race, religion, sexuality, mobility, disability, or education.</span></div>