M.Sc. Machine Learning student at KTH Royal Institute of Technology, with research interests in scientific machine learning, medical AI, agentic AI, computer vision, and probabilistic modelling.
I enjoy building machine learning systems that combine rigorous mathematics with practical applications, particularly in healthcare, combustion diagnostics, remote sensing, and trustworthy AI.
- Scientific Machine Learning
- Agentic AI & Retrieval-Augmented Generation
- Medical Image Analysis
- Computer Vision
- Probabilistic Machine Learning
- Meta-Learning
- Graph & Network Science
- Deep Learning for Healthcare
Agentic AI compliance assistant for Swedish SMEs using Google ADK and NVIDIA NIM, providing grounded responses from official government sources.
Machine learning framework for reconstructing combustion temperature profiles from laser absorption spectroscopy using Multi-Output Gaussian Processes.
Published research introducing a hybrid training framework for medical image registration by combining synthetic and real MRI data.
Publication: HybridMorph: Bridging the Gap between Synthetic and Real Data for Accurate MR Image Registration (NCVPRIPG 2025)
Deep learning methods for Polarimetric SAR image classification using complex-valued neural networks.
Implementation and comparison of gradient-based meta-learning algorithms for few-shot learning using PyTorch.
Languages
- Python
- C++
- MATLAB
- LaTeX
Frameworks
- PyTorch
- TensorFlow
- GPyTorch
- Google ADK
- scikit-learn
- OpenCV
Domains
- Machine Learning
- Computer Vision
- Medical Imaging
- Scientific Computing
- Reinforcement & Meta-Learning
- Bayesian Methods
I'm an active Linux user and strong supporter of open-source software.
My primary development environment is Fedora Linux, and I regularly work with openSUSE as well. Most of my research and development is carried out using open-source tools and libraries.
ð Website: https://caffineaddic.github.io
If you find something interestingâor something that can be improvedâfeel free to open an issue or start a discussion. Constructive feedback and collaboration are always welcome.
