The Euclid mission will deliver weak lensing data with unprecedented precision, which has the potential to revolutionise our understanding of dark energy and the growth of cosmic structures. Extracting its full information content requires going beyond the standard analyses. To make optimal use of the data, the OCAPi project will analyse Euclid's lensing maps directly at the pixel level. This approach, known as field-level inference, captures all the information and provides up to 5 times better constraints on the cosmological parameters (Porqueres et al. 2022, 2023).

This increased precision, however, requires an accurate modelling of the data. One of the main calibration challenges in weak lensing surveys is the redshift distribution of the lensed galaxies. Current calibration methods were designed for the standard analyses and may not be sufficiently accurate for field-level techniques. Quantifying the accuracy requirements and developing methods capable of reaching it is essential to enable field-level analyses of Euclid data and unlock the full scientific potential of the survey.

The goal of this PhD project is to develop a new redshift sampler for weak lensing, designed to meet the accuracy requirements of field-level inference. This sampler will combine physical models of galaxy populations with flexible machine-learning techniques. The thesis will contribute to maximising the potential of Euclid's weak lensing data and advance our understanding of the formation of cosmic structures.