Open projects for students, postdocs, and visitors working at the interface of artificial intelligence and astronomy.

A central theme across our team is learning the sky, or learning from the sky—building statistical and deep learning representations that learn directly from native telescope observations to clearly separate true astrophysical signals from instrument and atmospheric signatures.

You do not need to know both astronomy and artificial intelligence. Strong foundations in physics, signal processing, statistics, or computer science, paired with scientific integrity and curiosity about the other fields, is what we value most.

Areas where we are active and interested in new collaborations, student theses, or postdoctoral research:

• Multi-modal representations with sky surveys

  Learning shared continuous latent representations linking multi-band imaging, spectroscopy, and catalogues across wide-field surveys (including UNIONS, WEAVE, Euclid, Pristine, and Rubin/LSST). In the COSMIC-FM initiative, our goal is learning the sky by cleanly separating intrinsic astronomical signals from telescope optics, detector responses, and atmospheric transfer functions.

• Cosmological neural field inference

  Continuous field inference and uncertainty quantification directly on survey observations, from weak lensing shear and convergence maps to cosmic web matter density fields.

• Wide-field atmospheric correction with AI

  Mitigating atmospheric turbulence and reconstructing wide-field short-exposure observations (DanceCam) at extreme low photon counts, complementing traditional adaptive optics with data-driven wavefront and image restoration.

• Source separation and faint signal discovery

  Separating night-sky and telluric emission lines in multi-object spectrographs (e.g., WEAVE), foreground subtraction in 21cm radio cosmology, and detecting the faintest Trans-Neptunian Objects (TNOs) amidst dense background confusion.

• Pipelines for telescopes and surveys

  End-to-end data reduction, calibration, and quality-control pipelines for astronomical surveys—including WEAVE stellar spectroscopy, UNIONS weak lensing, Supernova surveys, and CFHT imaging. We do not only do AI: robust pipeline engineering and classical reduction methods are essential to all our science.

• Scientific computing and platforms

  Archive-scale distributed workflows and containerised environments on CANFAR and HPC systems, streaming multi-terabyte data cubes and enabling interactive analysis directly adjacent to archive storage.

• Postdocs

  NRC Herzberg postdoctoral programmes (Herzberg, Covington, or Plaskett fellowships) when open, or national fellowships (NSERC, CITA National, Banting) hosted jointly with a partner university. Please write at least two months before application deadlines so we can develop the research proposal together.

• Students

  Prospective graduate students should apply to UVic Physics and Astronomy, UVic Mathematics and Statistics, or UBC Computer Science, naming Seb Fabbro as a potential supervisor. Most graduate students are co-supervised with departmental faculty. For undergraduate and graduate internships, Canadian co-op programmes and direct thesis projects are available.

• Internships

  Mitacs Globalink Research Internships (GRI) and Accelerate programmes bring international students to Canada for research terms hosted through partner universities and NRC Herzberg.

• Collaborations

  We welcome shared research initiatives and visiting researchers from other universities and survey collaborations.

Funding depends on the year; reach out to discuss current availability.

Reach out if you would like to work together or just want to chat about an idea. A few helpful things to send along:

• What you are curious about or interested in working on
• A quick CV or a link to your code / GitHub / papers