The Department of Mathematics and Mathematical Statistics is offering a postdoctoral scholarship within the project “Geometry-Aware Autoencoders for Reliable Anomaly Detection”. The scholarship is full-time for two years, starting on 1 September 2026 or by agreement.
Department-specific information At the Department of Mathematics and Mathematical Statistics, we conduct research in both pure and applied mathematics. Research topics range from computational mathematics to the mathematical foundations of artificial intelligence. Within this environment, the research group in geometric deep learning is focused on developing the theoretical underpinnings of modern machine learning. We do this by using geometry to understand structured data, symmetries, and training dynamics, and by leveraging tools from, e.g., differential geometry to create novel models and analyse their properties. More information about our research can be found at https://www.umu.se/en/research/groups/geometric-deep-learning/.
Project description Autoencoders used for anomaly detection are central to modern data-driven discovery but are prone to poor performance when the problem has nontrivial geometric structure. An example of such a problem is the search for new physics beyond the standard model; the quest for a more complete understanding of the Universe. Experimental data is subject to geometric constraints arising from both physical laws and collider design, resulting in a detection problem that is as challenging as it is scientifically important.
The central goal of this project is to use neural ordinary differential equations (NODEs) on stratified spaces to create flow-based autoencoders that simultaneously accommodate the topology imposed by relativistic particle physics and the geometric constraints from particle detector systems. We will use tools from differential geometry to develop novel geometric deep learning models and explore their properties by training them on (simulated) data from real particle detectors (e.g., the LHC).
The project sits at the intersection of mathematics, physics and computer science, an area that is currently very active due to remarkable recent advances in scientific machine learning. The postdoctoral fellow will have the opportunity to participate in research in this exciting field in close collaboration with the geometric deep learning group at Umeå University. The appointed candidate will be able to hone their skills through access to experts in both differential geometry and particle physics and to develop their independence in a deep learning research environment with excellent opportunities for cross-disciplinary collaborations.
This postdoctoral scholarship is financed and administered by the Kempe Foundations (JCSMK261-0037). The stipend is tax-free and will be 750 000 SEK for two years, meaning 375 000 SEK per year.
Qualifications To qualify as a postdoctoral scholarship holder, the postdoctoral fellow is required to have completed a doctoral degree or a foreign degree deemed equivalent to a doctoral degree. This qualification requirement must be fulfilled no later than at the time of the decision about scholarship recipient.
Priority should be given to candidates who completed their doctoral degree, according to what is stipulated in the paragraph above, no later than three years prior. If there are special reasons, candidates who completed their doctoral degree before that may also be eligible. Special reasons include absence due to illness, parental leave, appointments of trust in trade union organisations, military service or similar circumstances, as well as clinical practice or other forms of appointment/assignment relevant to the subject area.
The doctoral degree should be in mathematics, physics, or computer science, or be considered to provide equivalent academic competence. A strong command of the English language, both written and spoken, is required, as are excellent communication and collaboration skills.
Good programming skills and documented experience implementing machine learning models and methods are required. Knowledge of differential geometry and differential equations, as well as previous research experience in geometric deep learning, is meritorious.
The successful candidate will demonstrate a commitment to continuously developing their skills and an ambition to contribute to the mathematical foundations of geometric deep learning.
Application A full application should include:
Cover letter in which you describe your research background and future ambitions, and how they align with the described project (maximum three pages),
Curriculum vitae (CV), including a publication list with a brief summary of your contribution to each publication,
Verified copy of doctoral degree certificate or documentation that clarifies when the degree of doctor is expected to be obtained,
Verified copies of other diplomas, lists of completed academic courses and grades,
Copy of doctoral thesis,
Other documents that the applicant wishes to claim,
Contact information for two people willing to act as references.
The application should be written in English or Swedish, and attached documents should be in Word or PDF format. The application should be registered via Umeå University’s e-recruitment system Varbi and submitted by the deadline 31 May 2026.
Umeå University strives to offer an equal environment where open dialogue between people with different backgrounds and perspectives lay the foundation for learning, creativity and development. We welcome people with different backgrounds and experiences to apply for this scholarship, and especially encourage female applicants.
More information Further details are provided by Associate Professor Fredrik Ohlsson, fredrik.ohlsson@umu.se.
