AI Engineer - Neural Scene Reconstruction

• Develop and improve neural scene reconstruction models using 3DGS, NeRF, and other learned 3D representations to reconstruct realistic driving environments.
• Implement dynamic scene reconstruction techniques, modeling static environments and moving traffic agents while handling occlusions, incomplete observations, and challenging environmental conditions.
• Develop neural rendering and novel-view synthesis methods to generate realistic camera observations from reconstructed 3D scenes.
• Contribute to multimodal sensor rendering, including camera, LiDAR, and radar simulation, to support realistic virtual testing environments.
• Build scalable reconstruction and rendering pipelines for efficient experimentation and large-scale processing of real-world driving datasets.
• Train and optimize reconstruction models, improving geometric accuracy, visual fidelity, temporal consistency, and computational efficiency.
• Develop evaluation tools and benchmarks to measure reconstruction quality, sensor fidelity, and real-to-sim consistency.
• Support interactive scene simulation, enabling viewpoint changes, scene editing, agent manipulation, and counterfactual scenario generation.
• Collaborate with simulation, perception, and autonomous driving teams to integrate reconstructed scenes into training, testing, and closed-loop evaluation workflows.
• Explore emerging research directions, including generative reconstruction, video diffusion, neural rendering, and world models.

Requirements

• 5+ years of practical experience in machine learning, computer vision, computer graphics, robotics, or related fields.
• Bachelor's or Master's degree in Computer Science, Computer Vision, Robotics, Electrical Engineering, or a related discipline, or equivalent practical experience.
• Solid understanding of 3D computer vision fundamentals, including multi-view geometry, camera models, coordinate transformations, and 3D representations.
• Familiarity with neural rendering or 3D reconstruction techniques, such as 3D Gaussian Splatting, NeRF, or related methods.
• Strong programming skills in Python and hands-on experience with deep learning frameworks, particularly PyTorch.
• Experience training, evaluating, and debugging deep learning models using real-world datasets.
• Familiarity with GPU-accelerated computing, differentiable rendering, or 3D vision libraries.
• Good analytical and problem-solving skills, with the ability to conduct systematic experiments and interpret model performance.
• Ability to collaborate effectively with researchers and engineers in a fast-paced R&D environment.

• Hands-on experience with 3D Gaussian Splatting, NeRF, or other neural scene reconstruction methods.
• Experience with dynamic or 4D reconstruction, moving object modeling, scene decomposition, or spatiotemporal representations.
• Familiarity with neural rendering frameworks such as gsplat, Nerfstudio, or similar tools.
• Knowledge of structure-from-motion, SLAM, camera calibration, pose estimation, or LiDAR point cloud processing.
• Experience with C++, CUDA, or GPU rendering optimization.
• Familiarity with generative models, video diffusion, or world models for scene reconstruction and novel-view synthesis.
• Knowledge of sensor simulation, including camera, LiDAR, or radar rendering.
• Experience with autonomous driving datasets, multimodal sensor data, or large-scale outdoor scene reconstruction.
• Familiarity with distributed training, large-scale data processing, or GPU computing infrastructure.
• Understanding of simulation fidelity, real-to-sim transfer, or closed-loop autonomous driving evaluation.
• Publications, open-source contributions, or research projects in computer vision, graphics, machine learning, robotics, or related fields.