Senior R&D Engineer

In this role, you will focus on the automated generation of roadside environments (buildings, street furniture, vegetation). You will combine state-of-the-art AI 3D Generation technologies (e.g., Trellis, TripoSR, Stable Fast 3D) with Procedural Content Generation (PCG) rules. Your goal is to build a pipeline that can automatically generate high-quality, diverse 3D assets from data or text prompts and procedurally assemble them into coherent, realistic driving scenarios along road networks.

Key Responsibilities

AI 3D Asset Pipeline (Trellis/GenAI) : Research and implement cutting-edge AI-to-3D pipelines. Utilize models like Trellis, TripoSR, or similar technologies to generate 3D meshes and textures for roadside buildings, barriers, and urban props from sparse inputs (text, images, or semantic labels).

PCG Scene Assembly : Design procedural algorithms to intelligently place AI-generated assets along the road network. This involves defining rules for density, spacing, and orientation to ensure that buildings and props (e.g., bus stops, streetlights, hydrants) are placed logically relative to the drivable area defined by HD Maps.

Automated Asset Optimization : Develop automated post-processing scripts to make AI-generated raw meshes game-ready. This includes auto-retopology, UV unwrapping, texture baking, and LOD (Level of Detail) generation to ensure the simulation runs efficiently in real-time.

Scene Variety & Variation : Leverage AI to create infinite variations of static objects (e.g., different architectural styles of shophouses, varied tree types) to solve the "repetition" problem in large-scale simulation worlds.

Requirements

Education: Bachelor's, Master's, or PhD in Computer Science, Computer Vision, Graphics, or related fields.

Experience: 5+ years in 3D Deep Learning, Procedural Generation, or Technical Art.

Core Technical Skills (Proficiency in at least one of the following)

AI 3D Generation : Hands‑on experience with 3D Generative models (e.g., Trellis, Gaussian Splatting, NeRF, Diffusion‑based 3D). Understanding of how to fine‑tune or deploy these models for specific asset classes (buildings vs. small props).

Procedural Content Generation (PCG) : Proficiency in Unity or Houdini (VEX/PDG). Ability to write logic for scattering and layout based on map data.

Geometric Processing : Strong knowledge of 3D mesh processing libraries (e.g., PyTorch3D, Trimesh, Blender Python API) for automating mesh cleanup and optimization.

Engineering & Coding :

Proficient in Python (for AI pipelines and asset processing) and C++ (for engine integration).

Experience with deep learning frameworks (PyTorch) and deploying models via Docker or cloud APIs.

Preferred Qualifications

Research Publications : Strong track record of publications in top‑tier computer vision or graphics conferences (e.g.,

CVPR, SIGGRAPH, ICCV, ECCV, NeurIPS ) related to 3D generation or neural rendering.

SOTA Implementation : Experience reproducing or improving upon recent papers like

Trellis ,

InstantMesh , or

Magic3D .

Real‑time Rendering : Understanding of PBR materials and real‑time lighting to ensure AI‑generated textures look correct in the engine.

Benefits

Competitive salary and benefits package.

Direct access to the latest compute clusters for training and inference of 3D Gen models.

Opportunity to pioneer the use of Generative AI in industrial‑scale simulation.

Note: Please include a portfolio or GitHub links demonstrating experience with AI 3D generation or procedural scene construction.

Fluent Chinese communication skills

enabling close collaboration with China‑based teams.

#J-18808-Ljbffr