Marek Oerlemans
Find out about my work
I’m Marek, and I work on deep learning for cancer risk prediction, with a focus on representation learning for pathology. I’m part of the AI for Oncology group at the Netherlands Cancer Institute — see aiforoncology.nl for more on our work.
Contact
If you have any interest or ideas relating to my work or my interests feel free to reach out to me.
| LinkedIn | Email | Google Scholar | | :—: | :—: | :—: |
Blog
Occasionally, if I find something interesting to write about such as my work as a PhD student or mathematical facts I found out I’ll post them here.
- On the equivalence of C-index and AUROC — February 2025
Selected Papers
-
S. Doyle*, M.A. Oerlemans*, et al. (2026). Enabling DCIS subtyping: leveraging foundation models for robust grading and molecular biomarker scoring. npj Breast Cancer. doi.org/10.1038/s41523-026-00957-6
* Shared first authors
-
M. Oerlemans, et al. (2026). Deep Multiple Instance Learning Predicts Gene Expression from Whole Slide Images in Ductal Carcinoma In Situ. Medical Imaging with Deep Learning (MIDL), Short Paper Track. openreview.net
Presentations
| When | Where | What |
|---|---|---|
| Jul 2026 | MIDL | Deep Multiple Instance Learning Predicts Gene Expression from Whole Slide Images in Ductal Carcinoma In Situ |
| Jan 2026 | NKI (spotlight talks) | Multimodal risk prediction for DCIS: combining H&E and gene expression |
| Jun 2025 | ECDP | Enabling robust biomarkers to stop overtreatment |
| Jun 2025 | Sanquin, bioinformatics meeting | Deep learning can see what pathologists cannot — predicting molecular biomarkers and grade in DCIS |
| May 2025 | Early Detection Conference | Deep learning can see what pathologists cannot — predicting molecular biomarkers and grade in DCIS |
| Nov 2023 | Cancer Grand Challenges, Future Leaders Conference | Predicting biomarkers for DCIS from WSI |
Experience
- PhD in Deep Learning — The Netherlands Cancer Institute, Amsterdam (2022–present, expected 2026). Developing deep learning methods for histopathology-based cancer recurrence risk prediction, spanning self-supervised learning, multiple instance learning, spatial point statistics, and high-dimensional regression. Organized the national PhD/PostDoc meeting in medical imaging; part of the admin team for a GPU cluster serving 100+ users across 64 GPUs.
- Deep Learning & Computer Vision Engineer — Nico.Lab, Amsterdam (2021–2022). Moved the registration framework for brain CT scans from an iterative algorithm to a deep-learning-based one.
- Graduate Intern, Research & Analytics — APG Asset Management, Amsterdam (2020). Solved a stochastic optimal control problem for portfolio rebalancing under price impact and translated it into trading rules for the portfolio management team.
Stack
- ML/DL: Python (PyTorch, scikit-learn, OpenCV), distributed training (DDP, FSDP), MLflow, C++/CUDA
- Infra: Linux, Slurm, Ansible
- CI/CD: Git, pytest, GitHub Actions, mypy, ruff
- Data: SQL, R
And Dutch, English, German, plus enough French and Italian to order coffee.
Education
| When | What and Where | Specifics |
|---|---|---|
| 2022 - present | PhD student at The Netherlands Cancer Institute and Leiden University. | Using deep learning on histology scans to do risk prediction in ductal carcinoma in situ. |
| 2020 - 2022 | MSc Applied Mathematics at Delft University of Technology. | Specifically focusing on numerical mathematics. |
| 2019 - 2021 | MSc Econometrics: Quantitative Finance at Erasmus University Rotterdam. | |
| 2016 - 2019 | BSc Econometrics and Operations Research at Erasmus University Rotterdam. |
Personal
Outside of work, I enjoy cycling, mountain biking, running, and rowing, and getting to the mountains to hike, climb, or ski. I live in Utrecht.