Batuhan Yildirim | CV Google Scholar Github Twitter LinkedIn


I'm a PhD candidate in the Molecular Engineering group at the University of Cambridge. I write software and research machine learning methods to facilitate the characterization and generation of atomic and nano structures.


Multi-task Scattering-Model Classification and Parameter Regression of Nanostructures from Small-Angle Scattering Data
Batuhan Yildirim, James Doutch and Jacqueline M. Cole
ACS Nano (2022). Submitted.
Calculating Small-Angle Scattering Intensity Functions from Electron Micrsocopy Images
Batuhan Yildirim, Adam Washington, James Doutch and Jacqueline M. Cole
RSC Adv. (2022)
[paper] [code]
Bayesian Particle Instance Segmentation for Electron Microscopy Image Quantification
Batuhan Yildirim and Jacqueline M. Cole
J. Chem. Inf. Model. (2021)
[paper] [code] [demo]
3-D Inorganic Crystal Structure Generation and Property Prediction via Representation Learning
Callum J. Court*, Batuhan Yildirim*, Apoorv Jain and Jacqueline M. Cole
J. Chem. Inf. Model. (2020)
*Equal Contribution
[paper] [code]
ImageDataExtractor: A Tool To Extract and Quantify Data from Microscopy Images
Karim T. Mukaddem, Edward J. Beard, Batuhan Yildirim, and Jacqueline M. Cole
J. Chem. Inf. Model. (2019)
[paper] [code] [site]

If you'd like to read any of my publications but can't get past the paywalls, feel free to email me and I'll send you a copy.


Spectral Graph Convolutions - Fourier Transforms (on Graphs) for the Birdwatcher

Open Source Software

imagedataextractor 2.0
a Python library for nanoparticle electron microscopy image quantification.
[site] [Github]
a Python library for fast computation of 2D and 3D radial distribution functions.
open source projects that I have contributed to include TorchEnsemble and ChemDataExtractor.


Electron Microscopy Particle Segmentation (EMPS) Dataset
a diverse collection of 465 electron micrographs and their corresponding particle instance segmentation maps.
[download] [Github]


I compete on Numerai, a data science tournament where I submit predictions weekly. I previously ranked in the top 0.1% and frequently rank in the top 10%. My best performing model currently is by256_ex_plus.
Generative Art
I enjoy creating generative art which exploits randomness, motion and colour to produce diverse sets of examples. I have sold many of my pieces and have created a Twitter bot that randomly generates and posts every 6 hours.

Contact: by256[at]