Post-Doctoral Research Visit F/M Learning imaging systems and solving inverse problems from unpaired data

Eighteen-month postdoctoral position in the Thoth team (computer vision and machine learning, Laboratoire Jean Kuntzmann, Université Grenoble Alpes), supervised by Julien Mairal. Inverse problems arise wherever images are acquired, from computational photography to microscopy, astrophysics and medical imaging: the quantity of interest is observed through a forward process (blur, noise, subsampling, sensor effects) and must be recovered from degraded measurements.

Plug-and-play solvers and diffusion- or flow-based generative priors work well when the forward operator is known, but in practice it rarely is, and calibration usually needs dedicated hardware or paired clean and degraded images. Recent Thoth work learned the forward operator from unpaired data by matching distributions on both the measurement and reconstruction sides through a differentiable plug-and-play solver. The postdoc will develop principled, scalable methods for learning imaging systems and solving inverse problems with scarce or no supervision, for example by extending unpaired operator learning to richer (spatially varying, nonlinear or unknown-noise) forward models, jointly learning operator and reconstruction with generative priors, studying identifiability, building models that adapt across devices, and applying them to computational photography or scientific imaging.

Start date 1 January 2027; gross salary €2,927 per month.

PhD in machine learning, computer vision, signal or image processing, applied mathematics or a related field with a strong publication record; good background in optimization, probability and statistics; strong Python and PyTorch skills; experience with inverse problems, generative models or optimal transport appreciated.