Leevi Kerkelä, PhD

Researcher & Software Engineer

Bio

I am a machine learning researcher and software engineer with academic training in physics and economics. I hold a PhD from University College London, where I subsequently worked as a research scientist, developing geometric deep learning methods for neuroimaging at UCL Computer Science. My professional experience spans physics research, neuroscience, finance, and software development. I am currently working as a founding AI Research Scientist at a venture-funded startup in stealth (coming soon). Outside of work, I enjoy board sports and time spent in natural environments.

Code

disimpy Python, Numba, CUDA
Massively parallel Monte Carlo diffusion MR simulator
noddi-python Python, Numba
Package for estimating NODDI parameters from diffusion MRI data
dexsnake Python, Ethereum, EVM
A Python package for trading on decentralised exchanges like Uniswap

Selected Publications

Human brain changes after first psilocybin use Lyons T, Spriggs M, Kerkelä L, Rosas FE, Roseman L, Mediano PAM, Timmermann C, Oestreich L, Pagni BA, Zeifman RJ, et al. Nature Communications, 2026
Rotation-invariant graph message passing enables acquisition protocol generalisation in learning-based brain microstructure estimation Kerkelä L, Zhang H arXiv, 2026
Spherical convolutional neural networks can improve brain microstructure estimation from diffusion MRI data Kerkelä L, Seunarine K, Szczepankiewicz F, Clark CA Frontiers in Neuroimaging, 2024
Example dataset of tensor-valued diffusion encoding MRI in the human brain Kerkelä L Zenodo, 2024
Comparative analysis of signal models for microscopic fractional anisotropy estimation using q-space trajectory encoding Kerkelä L, Nery F, Callaghan R, Zhou F, Gyori NG, Szczepankiewicz F, Palombo M, Parker GJM, Zhang H, Hall MG, et al. NeuroImage, 2021
Disimpy: A massively parallel Monte Carlo simulator for generating diffusion-weighted MRI data in Python Kerkelä L, Nery F, Hall MG, Clark CA Journal of Open Source Software, 2020

Contact

The best way to reach me is first name dot last name at protonmail dot com.