3D Reconstruction Algorithm Engineer_BCSC

Role Summary

Create the next generation data and simulation engine for autonomous driving, and explore the application of 3D Gaussian Splatting, neural rendering and generative AI in real road scene reconstruction, simulation video generation and closed-loop evaluation.

Participate in the research and development of complete links from real road test data reconstruction, controllable scene editing, dynamic target modeling, to simulation engine integration and autonomous driving algorithm verification, and promote 3DGS technology to truly serve the data production, scene generalization and closed-loop evaluation of ADAS/autonomous driving systems.

Design and implement an autonomous driving scene reconstruction algorithm based on 3D Gaussian Splatting / Neural Rendering;

Study multi-sensor reconstruction, camera trajectory optimization, geometric consistency, timing consistency and rendering quality improvement methods in large-scale road scenes;

Build a 3DGS simulation asset production link for LogSim/WorldSim to support high-fidelity reproduction and controllable editing of real road scenes;

Study the scene generalization capabilities such as dynamic target reconstruction and insertion, lane-level scene editing, weather/illumination/material changes, etc.;

Promote the integration of 3DGS results with simulation engines and autonomous driving perception/planning systems to achieve a data closed loop from reconstruction, rendering to closed-loop evaluation;

Optimize the efficiency of 3DGS training, rendering and inference, and support large-scale scene production and engineering deployment;

Collaborate with perception, planning, simulation, and data production teams to use reconstructed and generative simulation data for algorithm training, verification, and performance evaluation.

Requirements

Master's degree or above, majoring in computer science, artificial intelligence, electronic information, automation, surveying and mapping, robotics and other related majors;

• Familiar with deep learning and graphics related development tools such as PyTorch / CUDA / Python / C++;
• Have relevant R&D experience in at least one of 3D Gaussian Splatting, NeRF, Neural Rendering, 3D Reconstruction, SLAM, SfM, and MVS;
• Understand basic theories such as camera model, multi-view geometry, 3D reconstruction, coordinate transformation, pose optimization, and Bundle Adjustment;
• Applicants who are familiar with the data processing process of autonomous driving scenarios and can process multi-sensor data such as cameras, LiDAR, IMU, and GNSS will be given priority;
• Have strong engineering implementation capabilities and be able to transform algorithm prototypes into stable, large-scale production links;
• Have good problem analysis capabilities and be able to systematically optimize issues such as reconstruction quality, dynamic targets, cross-view consistency, and rendering artifacts.

Bonus points

• Familiar with 3DGS, Street Gaussian, PVG, EmerNeRF, UniSim, DriveDreamer, Cosmos, World Model and other related directions;
• Have project experience in autonomous driving simulation, LogSim, WorldSim, closed-loop simulation, SIL/HIL, Unreal Engine, Omniverse, Isaac Sim, etc.;
• Experience in road scene reconstruction, city-level reconstruction, dynamic scene reconstruction, and editable simulation asset production;
• Familiar with CUDA acceleration, real-time rendering, distributed training, model compression and inference optimization;
• Participated in open source projects, or published papers at conferences such as CVPR / ICCV / ECCV / NeurIPS / ICLR / ICRA / IROS / SIGGRAPH;

Have relevant experience in autonomous driving perception, BEV, Occupancy, 4D Label, data closed loop, corner case generation, etc.

About the Company

Bosch Innovation Software Development (Wuxi) Co., Ltd. provides future-oriented innovative technologies and cutting-edge solutions for the electrification, automation, interconnection, and personalization of transportation, and accelerates the realization and development of technology strategies for the Chinese market. The main development directions of the Bosch Software Center include intelligent connected cars, intelligent cockpits, assisted driving, vehicle-road-cloud collaboration, Industry 4.0, artificial intelligence big data, smart homes, and embedded software services. The Bosch China Innovation and Software Development Center empowers customers with software and has achieved innovative research and development results in many fields such as assisted driving, hydrogen fuel cells, heavy truck electric drive axles, and all-in-one controllers.