Director of Real2Sim Engineering

Role Overview

As Director of Real2Sim Engineering, you will own the architectural vision and execution of DreamVu’s simulation and synthetic data platform. Reporting directly to the CTO, you will bridge physical sensing hardware and AI/ML model deployment by building a state-of-the-art Real2Sim (and Sim2Real) pipeline built on the NVIDIA Omniverse / Isaac Sim ecosystem.

You will transition DreamVu to a simulation-first development model, enabling rapid foundation model training, automated synthetic dataset generation (SDG), and closed-loop evaluation on physical hardware.

Key Responsibilities

• Platform Architecture: Design and scale a GPU-accelerated Real2Sim architecture using NVIDIA Isaac Sim, Isaac Lab, and OpenUSD.

• Real2Sim & SDG Pipelines: Build pipelines that ingest real-world sensor streams (3D vision, spatial depth, point clouds) to auto-generate photorealistic digital twins and pixel-perfect annotated datasets.

• Sim2Real Transfer: Implement physics, material, and sensor-noise domain randomization to ensure zero/few-shot policy and vision transfer to physical hardware.

• Cross-Functional Execution: Partner with ML, Computer Vision, and Hardware teams to integrate simulation workflows into ROS 2 nodes, CI/CD pipelines, and cloud GPU clusters.

• Team Leadership: Recruit, lead, and mentor a high-performing team of simulation, graphics, and robotics engineers.

• QualificationsRequired

• 8+ years in robotics, physics simulation, 3D graphics, or computer vision, including 3+ years in technical leadership.

• NVIDIA Ecosystem Expertise: Production experience architecting workflows in NVIDIA Isaac Sim, Isaac Lab, or Omniverse.

• OpenUSD Foundations: Strong mastery of Universal Scene Description (OpenUSD) schema, composition arcs, asset variants, and physics definitions.

• Software Stack: High proficiency in Python (Isaac Sim API, PyTorch/ML workflows) and C++.

• Sim2Real Transfer: Demonstrated success using domain randomization, synthetic data generation, and sensor modeling to train models deployed on physical robots/devices.

Preferred

• Experience with 3D neural reconstruction (3D Gaussian Splatting, NeRFs, RGB-D mesh generation).

• Background in real-time game engines (Unreal Engine 5 / Unity C++) prior to OpenUSD adoption.

• Exposure to generative world models (e.g., NVIDIA Cosmos, GR00T foundation models).

• Experience with cloud GPU orchestration (AWS/GCP, Docker, Kubernetes) for large-scale parallel simulations.