Simulation Engineer

Full Time

|

Philadelphia, PA

|

DreamVu AI

Simulation Engineer

DreamVu

 | Full-Time | Philadelphia/ Hyderabad/ Hybrid



About DreamVu

DreamVu provides training data infrastructure for humanoid robots that actually work. We solve the fundamental challenge of physical AI: creating complete multimodal datasets that enable humanoid robots to learn manipulation, coordination, and skill transfer in real-world environments.


Our end-to-end platform transforms real-world captures into ready-to-train data by pairing standard egocentric cameras with our patented Alia 360° camera system. This delivers the complete picture humanoids need—vision, language, and action data from both first-person and third-person perspectives—enabling teams to build robots that perform reliably outside the lab.



The Opportunity

We're seeking a highly skilled 

Simulation Engineer

 to play a critical role in our simulation-to-real pipeline. This is a high-impact position that bridges 3D artistry with robotics engineering, directly supporting our mission to close the sim-to-real gap for humanoid robot training.


You'll be responsible for transforming real-world 3D captures and CAD models into production-ready, physics-accurate simulation assets for NVIDIA Isaac Sim. This isn't a traditional 3D modeling role—it demands deep technical expertise in USD pipelines, physics simulation, and Python automation to support our digital twin generation and synthetic data creation workflows.



What You'll Do

Asset Development & Optimization


  • Process 3D reconstructions from our 360° capture system and convert them into simulation-ready assets

  • Configure assets for multimodal AI training: semantic labeling, instance segmentation, 6DOF pose tracking, and manipulation affordances

  • Design and implement collision geometries, articulation systems, and physics properties (mass, friction, contact) for realistic humanoid-environment interaction

  • Optimize assets for real-time performance using LODs, proxy meshes, and efficient material workflows while maintaining photorealistic quality


Digital Twin Environment Construction


  • Build and validate photorealistic simulation scenes that mirror our real-world capture environments

  • Ensure physics determinism and computational efficiency for large-scale synthetic data generation

  • Create modular, reusable scene components that preserve multimodal annotations across vision, language, and action data streams

  • Maintain consistency between physical capture spaces and their digital twin counterparts


Pipeline Development & Automation


  • Write Python scripts and OmniKit extensions to automate the 3D-to-simulation asset processing workflow

  • Develop tools for batch processing hundreds of captured environments into Isaac Sim-compatible scenes

  • Integrate with our annotation pipeline to ensure semantic labels, trajectories, and skill demonstrations transfer correctly

  • Document repeatable pipelines that enable rapid scaling across diverse humanoid training scenarios


What You Bring

Required Qualifications


  • NVIDIA Isaac Sim & Omniverse

    : Demonstrated expertise with the platform, toolset, and simulation workflow

  • USD (Universal Scene Description)

    : Strong command of layering, variants, composition arcs, and USD best practices for complex scene assembly

  • Physics Simulation

    : Proven experience creating accurate colliders (convex decomposition, primitive fitting), configuring articulated joints, and tuning material properties for contact dynamics

  • Python Scripting

    : Ability to write production-quality scripts for Isaac Sim extensions, automated asset processing, and workflow integration

  • 3D DCC Proficiency

    : Skilled in Blender, Maya, or 3ds Max for photogrammetry cleanup, UV optimization, and geometry preparation

Preferred Qualifications


  • Experience with photogrammetry reconstruction tools (RealityCapture, Metashape, NeRF/Gaussian Splatting)

  • Background in robotics simulation, humanoid kinematics, or manipulation task environments

  • Understanding of synthetic data generation, domain randomization, and sim-to-real transfer concepts

  • Familiarity with physical AI training data formats (RLDS, HDF5) and multimodal annotation workflows

  • Experience with Unreal Engine or other real-time rendering pipelines

  • Knowledge of vision-language-action (VLA) model requirements and training data structures


Why DreamVu

  • Work at the cutting edge of physical AI and humanoid robotics with technology that solves real sim-to-real challenges

  • Direct impact on training data infrastructure used by leading humanoid development programs

  • Collaborate with experts in computer vision, robotics, and machine learning

  • Help build the datasets that will enable the next generation of capable humanoid robots

  • Competitive compensation and comprehensive benefits package

  • Opportunity to grow with a company solving one of the hardest problems in robotics


Ready to Apply?

If you're passionate about building the simulation infrastructure that enables humanoid robots to learn and perform in the real world, we want to hear from you. Send your resume, portfolio showcasing relevant simulation and asset work, and a brief note about what excites you about this role to careers@dreamvu.com