Multiprobe frameworks

End-to-end, differentiable pipelines combining CMB and LSS probes for robust cosmology.

During my PhD, I built an end-to-end, differentiable pipeline that combines multiple cosmological probes—weak lensing, galaxy clustering, BAO, CMB primary anisotropies, integrated Sachs-Wolfe effect, and CMB lensing—in a single, self-consistent inference framework. This includes simulation-based covariances, fast emulators, and cross-correlation measurements to maximise information while controlling systematics across datasets.

The pipeline was first developed and validated on mock data in Reeves et al. (2024), then applied to real data in Reeves et al. (2025a), where we explored neutrino mass constraints and dynamical dark energy. Most recently, the framework has been extended to DESI Legacy Imaging Survey data in Reeves et al. (2025b), where we compared early and late-time dark energy models.

Pipeline overview

Multiprobe pipeline architecture: Schematic showing the end-to-end framework combining simulations, theory predictions, and observational data across multiple cosmological probes. The pipeline uses simulation-based covariances and fast emulators to enable joint inference across CMB and LSS datasets.

Subset of pipeline spectra

Measured data vector: Cross-correlations between DESI Legacy Imaging Survey (galaxy positions and shapes) with CMB lensing and the integrated Sachs-Wolfe (ISW) effect. The theory prediction (red) shows excellent agreement with the measured data (black points). From Reeves et al. (2025b).