This position is in the context of our ongoing ERC Advanced Grant NERPHYS (for more information see https://www.inria.fr/en/erc-grants-george-drettakis-ai-physics-3d and https://project.inria.fr/nerphys/)
The are seeking to hire a highly motivated candidate who will be part of this exciting project, involving software engineering work on novel research projects, while working alongside the GRAPHDECO group that is continuing the advancement of new research ideas in this project.
This position is a great opportunity to be part of a world-class team of researchers working on exciting and timely projects. The successful candidate will acquire top-notch first-hand knowledge and experience in radiance field rendering which is in extremely high demand today, providing excellent skills for career enhancement.
The engineer will work on projects of the Ph.D. students and postdocs of the group.
Generative models have proven highly effective at providing powerful structural and physical priors for complex computer graphics and vision tasks. Recent breakthroughs leverage these generative priors to tackle inverse problems, such as intrinsic image decomposition [1], radiance field relighting [2], and physical dynamic reconstruction.
Most state-of-the-art approaches build on large video or multi-view diffusion models pre-trained on massive datasets. While fine-tuning these models preserves generalization while adapting them to domain-specific physics and rendering constraints, doing so reliably requires robust data pipelines, dynamic simulation grounding, and large-scale data collection strategies. Within the ERC Advanced Grant NERPHYS (NEural Representations for PHYSical simulation), the engineer will help build solutions that aim to bridge neural representations (e.g., 3D Gaussian Splatting, NeRFs) with physics-based simulation and generative priors to achieve interactive, physically plausible 3D scene editing and simulation.
