Software Simulation Engineer, Sensor Rendering

Basis Set is focused on building the future of embodied intelligence through AI-driven systems. They are seeking a Software Simulation Engineer to develop and scale sensor simulation infrastructure for sim-to-real model training, producing photorealistic synthetic data for perception systems.

Responsibilities:

• Implement and validate physics-based sensor simulation models (structured light, depth, RGB, stereo, etc.) within platforms such as NVIDIA Isaac Sim, Blender or Unreal Engine, producing outputs that closely match real sensor behavior
• Build photorealistic scene rendering pipelines that account for sensor placement on the robot end-effector. Emulate robot trajectories both with and without physical models. Utilize accurate material properties, such as metal reflectance, weld spatter, and torch glow, to ensure synthetic data is meaningful for perception model training
• Develop synthetic data generation pipelines producing annotated ground-truth (point clouds, depth maps, segmentation masks) at scale
• Implement domain randomization strategies (lighting, material variation, sensor noise, viewpoint perturbation) to improve sim-to-real transfer for downstream perception models
• Collaborate with perception teams to ensure rendered outputs meet dataset requirements and write high-quality Python code
• Lead the design and validation of high-fidelity, photorealistic sensor render pipelines grounded in real sensor characterization data and validated against physical measurements
• Architect the sensor rendering strategy for the Perception team, defining which sensor modalities, material models, and environmental conditions must be simulated to support perception across the full weld cell workflow
• Own the sim-to-real validation framework: define quantitative benchmarks and go/no-go criteria for when synthetic sensor data is ready to feed production model training
• Drive 3D asset and environment pipeline strategy, including CAD-to-simulation workflows, SDF/URDF asset management, material library management, and procedural scene generation for weld cell environments across Gazebo, Isaac Sim, and Unreal Engine
• Define strategy for when and how to use each simulation platform (Gazebo for ROS-integrated functional testing, Isaac Sim or Unreal Engine for photorealistic synthetic data generation) and build workflows that span them coherently
• Mentor engineers on rendering best practices, physically based material modeling, Gazebo plugin development, and synthetic data methodology