Professional Experience
AI Researcher in Digital Health
Dec 2024 –
PresentEurecat, Digital Health Unit Barcelona, Spain
- Develop generative models (pixel-space and latent diffusion) that
synthesise medical images, mainly 3D chest CT, conditioned on anatomy
and clinical metadata to create new training data and forecast tumour
growth.
- Built LAND, an anatomically guided latent diffusion model for
high-resolution 3D chest CT (Scientific Reports, 2026).
- Built MUMINS, which predicts a patient’s follow-up scan from a
baseline scan with a per-voxel uncertainty map, validated on lung CT and
brain MRI (WACV 2027).
- Trained 3D models both on low-memory GPUs (resource-constrained)
and multi-GPU on HPC clusters.
Computer Vision Engineer
Nov 2023 – Nov
2024Intek Medical Barcelona, Spain
- Developed AI software for automated hair and skin analysis,
supporting personalised, patient-specific treatments.
- Applied deep learning to 3D human reconstruction to obtain
anthropometric measurements.
- Designed AI-assisted data acquisition and annotation tools, used
to build a large dataset for model training.
Research Intern
May 2023 – Sep
2023Computer Vision Center (CVC) Barcelona, Spain
- Applied Large Language Models for Document Visual Question
Answering using PyTorch and Hugging Face.
- Improved performance by adding spatial layout information
(bounding boxes) to the prompt.
- Added the visual modality through a Vision Transformer so the
language model could interpret visual tokens.
Research Intern
Jun 2022 – Aug
2022LTS5, EPFL Lausanne, Switzerland
- Worked on the “3T-study” (Test, Triage and Treat) cervical cancer
screening programme led by the Geneva University Hospitals (HUG) and the
Dschang District Hospital, Cameroon.
- Developed deep learning models to detect strong motion in
smartphone videos recorded during VIA screening.
Publications
MUMINS: Metadata-conditioned Uncertainty-aware Medical Image
Next-state Synthesis
2027A. Oliveras, R. Marí, R. Redondo, O.
Guardià-Olivella, C. I. Ugwu, A. Tost, B. Nagarajan, C. Migliorelli, V.
Ribas, P. Radeva
IEEE/CVF Winter Conference on Applications of Computer Vision
(WACV 2027), accepted | arXiv:2609.17169 | Code
Anatomically Guided Latent Diffusion for High-Resolution 3D
Chest CT Synthesis
2026A. Oliveras, R. Marí, R. Redondo, O. Guardià, C. I.
Ugwu, A. Tost, B. Nagarajan, C. Migliorelli, V. Ribas, P. Radeva
Scientific Reports (Nature Portfolio), 2026 | doi:10.1038/s41598-026-51634-4
| Project
page
Technical Skills
Machine learning: Diffusion models, generative
models, uncertainty estimation, foundation model training and
fine-tuning, Vision Transformers, LLMs and vision-language models
Training at scale: PyTorch DDP, FSDP, mixed
precision, gradient checkpointing, PyTorch Lightning
Medical imaging: 3D CT and MRI, DICOM, NIfTI, MONAI,
SimpleITK, nibabel, 3D Slicer
HPC and MLOps: Slurm, Docker, Singularity, Linux,
Weights & Biases, TensorBoard, MLflow
Programming: Python, C/C++, Java, MATLAB, Shell
scripting, HTML5/CSS
Libraries and tools: PyTorch, Hugging Face,
TensorFlow, Keras, OpenCV, Scikit-Learn, NumPy, Pandas, PyQt,
Git/GitHub, Blender, MeshLab
Data and cloud: SQL, MongoDB, Microsoft Azure