Software Engineering Intern, Neural Reconstruction and Rendering - 2027

Location: China, Shanghai

Time Type: Full time

Job Description

Join NVIDIA’s Real2Sim team to turn real-world captures into accurate, editable simulation environments for robotics. You will develop and evaluate reconstruction and rendering methods that improve scene geometry, visual fidelity, and computational efficiency, contributing to workflows built around NVIDIA NuRec and Isaac Sim.

What you’ll be doing

• Improve neural scene reconstruction through Gaussian optimization, sampling, and geometry-aware learning, addressing incomplete surfaces, artifacts, and challenging viewpoints.
• Explore surface extraction, material estimation, relighting, and generative refinement to make reconstructed environments more accurate and editable.
• Implement and optimize research methods in reusable reconstruction and rendering pipelines, collaborating with engineers to integrate results into NuRec and Isaac Sim workflows.
• Build reproducible experiments that measure geometric accuracy, novel-view quality, consistency across views and time, runtime, and GPU memory, and evaluate the impact on a selected robotics simulation use case.

What we need to see

• Pursuing a master’s or doctoral degree in computer science, computer graphics, computer vision, robotics, or a related field.
• Strong Python and PyTorch skills, with experience implementing, training, and debugging research models on GPUs.
• Hands-on research or project experience in 3D Gaussian Splatting, neural rendering, multi-view reconstruction, or closely related methods.
• Solid foundations in 3D geometry, camera models, numerical optimization, and machine learning.
• Ability to reproduce research results, design controlled comparisons, analyze failure cases, and collaborate effectively with researchers and engineers.

Ways to stand out from the crowd

• Depth in Gaussian sampling and densification, including MCMC-based optimization; surface reconstruction and meshing; or inverse rendering and relighting.
• Experience with generative priors for reconstruction, scene completion, or rendering, including preserving geometry and consistency across views.
• Experience with C++, CUDA, Slang, differentiable rendering, or GPU performance profiling.
• Familiarity with gsplat, 3DGUT/3DGRT, NuRec, OpenUSD, or Isaac Sim; effective use of AI-assisted development with code review and validation.