Staff Data Scientist – Simulation Motion AI

Intuitive•San Francisco, CA

About The Position

As a Staff Data Scientist – Simulation Motion AI, you will lead the design of scalable simulation-data systems used to train, evaluate, and improve robotic policies for complex, safety-critical motion AI products. You will work alongside foundation and policy model teams, test, regulatory, product and clinical teams to generate data, train and test algorithms in simulation at scale for motion AI products.

Requirements

PhD or Master’s degree in Computer Science, robotics, machine learning, data science, electrical engineering, mechanical engineering, applied mathematics, or a related technical field.
9+ years of industry experience, post training, developing simulation, machine learning, robotics, computer vision, simulation, or autonomous-systems software; or 4+ years of industry experience with a PhD
Proven hands-on experience with at least one major robotics simulation platform: NVIDIA Isaac Sim, NVIDIA Isaac Lab, MuJoCo or Equivalent physics-based robotics simulation environment
Expertise in creating custom simulation environments, robot assets, task definitions, sensors, reward functions, termination criteria, or procedural scene-generation systems.
Expertise with Python and modern machine-learning frameworks such as PyTorch
Experience working with robot kinematics, dynamics, coordinate frames, calibration, trajectory representations, and closed-loop control and designing datasets and experiments for multimodal or time-series machine learning.
Demonstrated ability to diagnose model failures using quantitative analysis rather than relying solely on aggregate success metrics.
Experience deploying, testing, or validating models on physical robotic systems.

Nice To Haves

Experience developing vision-based manipulation policies using RGB, stereo, depth, segmentation, optical flow, keypoints, or learned visual representations.
Experience with transformer, diffusion, or vision-language-action for robotics.
Experience with deformable-object simulation, articulated objects, fluids, cables, sutures, tissue, or other contact-rich environments.
Experience with experiment tracking, dataset versioning, distributed training, and large-scale simulation infrastructure.
Familiarity with containerized and distributed computing technologies such as Docker, Kubernetes, Slurm, Ray, or cloud GPU platforms.
Experience in a safety-critical field such as medical robotics, autonomous vehicles, aviation, industrial automation, or defense.
Proficiency in GPU optimization for either inference or rendering

Responsibilities

Architect scalable pipelines for generating synthetic and procedurally varied robotic interaction data using Isaac Sim, Isaac Lab, MuJoCo, or similar simulators.
Build SOTA, scalable simulation environments developing task data generation, ML, and scene domain randomization strategies covering: Robot initial conditions and calibration errors, Camera poses, optics, lighting, and occlusions, Object geometry, appearance, pose, and material properties, Contact, friction, compliance, deformation, and force variation, Tool wear, latency, noise, and actuator uncertainty, Nominal, edge-case, failure, and recovery conditions.
Define data-generation curricula that progress from constrained primitive actions to long-horizon, multi-stage robotic tasks.
Generate targeted recovery data for rare but safety-relevant conditions, including failed grasps, object displacement, camera obstruction, tracking loss, unexpected contact, and partial task completion.
Design systems for efficiently replaying, perturbing, relabeling, and extending real-world robot trajectories in simulation.
Establish provenance, metadata, versioning, and reproducibility standards for simulated datasets.