• What is causing cosmic acceleration?

  If dark energy evolves with time, or gravity departs from general relativity on large scales, the evidence has to agree across expansion history, the growth of structure and weak lensing at the same time. We work on this with UNIONS, Euclid and LSST data.

• Did inflation involve more than one field?

  Primordial non-Gaussianity leaves a scale-dependent imprint on how galaxies cluster on the largest scales, an imprint that foregrounds and calibration residuals can mimic. Separating the two needs tracers that share the cosmology but not the systematics.

• What do the oldest stars remember?

  The most metal-poor stars in the Milky Way carry the chemistry of the first supernovae, and their orbits trace how the halo was assembled. Pristine photometry, Gaia astrometry and high-resolution spectroscopic programs measure the same stars in independent ways.

• Learning a sky model from incomplete views

  There is no ground truth in astronomy. Images have no clean version, labelled samples are small and biased toward bright objects, and simulations assume the physics under test. We build models that recover what noisy, partial observations share without erasing faint or rare signals.

• One sky across instruments

  Every survey sees the sky through its own resolution, passbands, depth, cadence and selection. Instead of one catalogue per survey, we fit a single continuous model of the sky to each instrument’s own measurements, so faint and blended sources are constrained by all the data at once.

• Predictions that survive new data

  A prediction is only useful where the data support it. We develop calibrated predictive distributions and conformal methods that flag when a model extrapolates, for example from bright spectroscopic training sets to faint photometric targets, and we use independent probes as the external check.

We work directly with wide-field survey data and archives: UNIONS (CFHT, Subaru, PanSTARRS), WEAVE, Euclid, Vera C. Rubin Observatory (LSST), Pristine, and the Canadian Astronomy Data Centre (CADC).

Our technical work also includes pipeline engineering for telescopes and surveys—such as WEAVE stellar spectroscopy, UNIONS weak lensing, Supernova surveys, and CFHT imaging.