thesis-desi-cmb-fli
Repository scaffold for a thesis project targeting field-level inference (FLI) on DESI galaxy samples cross-correlated with CMB lensing reconstructions from Planck and ACT. The scientific pipeline itself remains unimplemented; the supporting infrastructure has been preserved so the cosmological workflow can be rebuilt cleanly.
Thesis vision
- Construction of a reproducible FLI workflow that ingests DESI galaxy clustering data and CMB lensing maps (Planck PR4, ACT DR6+).
- Delivery of joint cosmological constraints from the cross-correlation of those observables, with clear validation and documentation.
- Maintenance of publication-grade software practices (tests, docs, metadata) throughout thesis development.
What is already in place
- Python package stub (
desi_cmb_fli) exposing version metadata. - Automation hooks: MkDocs-based documentation, pytest configuration, formatting
via
ruff. - Repository structure for configs (
configs/), generated figures (figures/), and documentation (docs/), ready to host the real pipeline components. - GitHub Actions for continuous testing and doc deployment—pushes trigger the test suite and documentation builds rerun automatically when relevant content changes.
Quick Start
Local development (CPU only):
conda env create -f env/environment.yml
conda activate desi-cmb-fli
pip install -e .
pre-commit install
NERSC Perlmutter (with GPU): See docs/hpc.md for complete setup.
Development
Run tests: pytest (DESI tests auto-skip on non-NERSC systems)
Format code: ruff format .
Preview docs: mkdocs serve
Git hooks automatically format code on commit. CI runs tests on push.
Building the cosmology analysis
- DESI/Planck/ACT interfaces and FLI likelihood modules are intended to reside
under
src/desi_cmb_fli/. - Experiment configurations (tracer selections, mask definitions, inference
grids) should be tracked in
configs/with lightweight examples committed to the repository. - The
tests/directory is expected to grow with unit, integration, and regression checks covering the forward model, likelihood evaluations, and validation diagnostics. - Scientific logic should be documented in
docs/, including dataset handling, field modeling choices, validation strategies, and interpretation of cosmological posteriors.
Citation
The project scaffold for the FLI × DESI × CMB lensing analysis should be cited
using CITATION.cff.
License
MIT (see LICENSE).