Research Scientist, Neural Reconstruction

Waabi World depends on accurate, scalable, and efficient reconstruction of the 3D/4D physical world from real-world sensor data. As a Research Scientist in Neural Reconstruction, you will develop the next generation of neural scene-representation and reconstruction algorithms that transform sparse, noisy, and partially observed driving data into realistic and controllable digital worlds.

This role focuses on 3D/4D neural reconstruction and rendering, including 3DGS/NeRF, neural scene representation, and generalizable reconstruction models. Your work will directly power Waabi World’s ability to build high-fidelity scene assets, recover geometry and dynamics from sensor data, and support realistic simulation and rendering at scale.

 

You will…

- Conduct fundamental and applied research in neural reconstruction, including:

• 3DGS / NeRF

• Dynamic scene reconstruction

• Feed-forward reconstruction

• Multi-sensor scene representation learning 

- Build scalable reconstruction and simulation systems for dynamic urban scenes, including vehicles, background, lighting, and long-range structure.

- Collaborate with simulation engineers to integrate models into large-scale, distributed training and rendering pipelines.

- Publish high-impact research at top conferences (CVPR, ECCV, ICCV, NeurIPS, ICLR, ICRA, SIGGRAPH).

- Mentor junior scientists and interns; foster a culture of scientific rigor and rapid experimentation.

- Stay current with emerging advances in neural rendering, 3D representation learning, differentiable rendering, and scalable reconstruction systems.

 

Qualifications:

- Demonstrated technical innovation:  You have a Ph.D. in Computer Vision, Machine Learning, Robotics, or a related field or equivalent research experience pushing the boundaries of a technical field..

- Strong prototyping and implementation:  You have expert-level Python & PyTorch (or JAX) skills; strong software-engineering fundamentals and experience with distributed training.

- Expert domain knowledge: You have built  generative or predictive models of the physical world with scale and efficiency in mind for real-world applications

- Team player:  You have worked in a close-knit team of researchers and engineers and have strong communication  to deliver successful projects. 

 

Bonus:

- Proven ability to translate research into production-quality code and measurable product impact.

- Demonstrated first-author publications in top-tier venues on topics such as:

• 3DGS / NeRF / neural rendering

• 3D / 4D reconstruction

• Generalizable reconstruction

• Geometry-aware or multi-sensor representation learning

- Experience working with camera, LiDAR, maps, and large-scale driving datasets.

- Strong background in graphics, geometry, rendering systems, or simulation.