PhD Scholarship: Error Propagation and Implicit Priors in Computational UQ

About the role

Three-year fully funded PhD at DTU Compute’s Section for Scientific Computing, part of the CUQI (Computational Uncertainty Quantification for Inverse problems) research initiative funded by the Independent Research Fund Denmark. The project examines how data errors propagate in iterative regularisation methods such as CGLS and GMRES, with applications to computed tomography and image deblurring from dental scanners and fusion tomography. Results are implemented in the open-source CUQIpy Python package. Start date: 1 September 2026. Deadline: 10 June 2026.

Responsibilities

• Analyse error propagation in iterative Krylov-subspace solvers (CGLS, GMRES) for ill-posed inverse problems.
• Develop implicit-prior models and assess how they influence uncertainty estimates in CT reconstruction.
• Implement methods in CUQIpy and validate against real dental and fusion tomography datasets.
• Complete 30 ECTS of advanced coursework during the programme.
• Support teaching and project supervision at DTU Compute.

Requirements

• Two-year Master’s degree (120 ECTS) in Applied Mathematics, Computational Mathematics, or an equivalent quantitative discipline, completed before the start date.
• Strong grounding in numerical linear algebra, regularisation theory, or statistical inverse problems.
• Proficiency in Python.
• Fluent English, written and spoken.

Nice to have

• Prior experience with Bayesian inference or probabilistic programming frameworks.
• Familiarity with tomographic reconstruction (ASTRA Toolbox, ODL, or similar).
• Open-source software contribution history.

How to apply

Apply through the DTU career portal using the link above. Upload your CV, cover letter, degree transcripts, and details for two academic references. Salary follows the collective agreement of the Danish Confederation of Professional Associations. Deadline is 10 June 2026; start date is 1 September 2026.