I started at AstraZeneca in November 2018 because I want to help cure diseases by discovering new drugs. In my role, I apply machine learning and neural network technology to speed up and augment the drug discovery process.

Before I joined I worked as a consultant for various pharma companies for several years, intersected with two postdoc projects. During my last years as a consultant I also conducted research on how to apply novel neural network architectures for generative modelling and deep learning quantitative structure–activity relationship (QSAR) models.

Within AstraZeneca Discovery Sciences, I get the opportunity to continue my scientific research into use of deep learning and neural networks for drug discovery. My focus is to drive the science and method development behind use of AI and machine learning in drug discovery.

Here it’s possible to get immediate feedback from projects and see the methods provide value. I get knowledge about how the algorithms perform in real life, which leads to new and relevant ideas for further research and improvements.

AstraZeneca is a great place to work when it comes to Data Science and AI because we are not just talking about it, we are using it and developing cutting edge methods hoping to to ultimately improve patients’ lives.

Esben Jannik Bjerrum Principal Scientist, Cheminformatics, Discovery Sciences, R&D


2003: H.C. Ørsted bronze medal for distinguished studies


Principal Scientist, Cheminformatics, Discovery Sciences, R&D


CEO, Molecular Machine Learning and Data Science Consultant, Wildcard Pharmaceutical Consulting


Ph.D., Computational Chemistry

  Featured publications

SMILES enumeration as data augmentation for neural network modeling of molecules.

Bjerrum, EJ. arXiv preprint arXiv:1703.07076 (2017).

Molecular generation with recurrent neural networks (RNNs).

Bjerrum, EJ, and Threlfall R. arXiv preprint arXiv:1705.04612 (2017).

Data Augmentation of Spectral Data for Convolutional Neural Network (CNN) Based Deep Chemometrics.

Bjerrum EJ, Glahder M, and Skov T. arXiv preprint arXiv:1710.01927 (2017).

De Novo Molecular Design by Combining Deep Autoencoder Recurrent Neural Networks with Generative Topographic Mapping.

Sattarov B, et al. Journal of chemical information and modeling (2019).

Improving Chemical Autoencoder Latent Space and Molecular De Novo Generation Diversity with Heteroencoders.

Bjerrum, E and Sattarov B. Biomolecules 8.4 (2018): 131.