三维重建算法工程师_BCSC
Core
Develop next-generation data and simulation engines for autonomous driving using 3D Gaussian Splatting, neural rendering, and generative AI for real-world scene reconstruction and closed-loop evaluation.
Role type
Senior IC 3D reconstruction and neural rendering algorithm engineer
Builds
High-fidelity 3D Gaussian Splatting simulation assets and reconstruction pipelines for LogSim/WorldSim
Domain
Autonomous driving / 3D computer vision / Neural rendering
Deliverable
production ML models
Required skills
3D Gaussian Splatting, Neural Rendering, 3D Reconstruction, SLAM, SfM, MVS, Camera models, Multi-view geometry, Pose optimization, Bundle Adjustment, Multi-sensor data processing (Camera, LiDAR, IMU, GNSS), CUDA, PyTorch, C++, Python
Preferred skills
Street Gaussian, PVG, EmerNeRF, UniSim, DriveDreamer, Cosmos, World Model, Unreal Engine, Omniverse, Isaac Sim, Distributed training, Model compression, Real-time rendering, Open source contributions, Top-tier conference publications (CVPR/ICCV/ECCV/NeurIPS/ICRA/IROS/SIGGRAPH), BEV, Occupancy, 4D Label, Corner case generation
Responsibilities
Design and implement 3DGS/Neural Rendering algorithms for autonomous driving scene reconstruction; Research multi-sensor reconstruction and rendering quality improvement in large-scale road scenes; Build 3DGS asset production pipelines for LogSim/WorldSim; Research dynamic object reconstruction and scene generalization (weather/lighting/material changes); Integrate 3DGS results with simulation engines and ADAS/perception/planning systems; Optimize training, rendering, and inference efficiency for large-scale production; Collaborate with perception, planning, and simulation teams to validate algorithms using generated data.
