Carl Nettelblad

Short presentation

My research is in the field of scientific computing, but with a firm focus on life science application data analysis, utilizing modern computing architectures (including GPU computations and massive paralellism in varying forms). My basic question is "how can we trade experiment result quality for more sophisticated computational methods", giving better results with worse original data.


  • artificial intelligence
  • genomics
  • hidden markov models
  • hpc
  • machine learning
  • optimization
  • xfel


The advances over the last two decades in techniques and methods for massive dat collection have been tremendous in many sectors, including life science. The technological development have allowed ever larger data sets. The sizes being analyzed easily surpass the point where a single scientist can perform any kind of thorough manual quality control. Therefore, the development of analysis methods where errors and inaccuracies can be automatically identified and handled, is crucial.

A central aspect for my research is thus that primary data will always contain errors, missing data, and noise. Based on this insight, one can try to develop metods for handling "bad" samples, or to allow less expensive experimental methodology for equivalent results. This can be contrasted against the more established approaches, which basically imply filtering heavily to identify high-quality sections in datasets, or to use methods that were really designed for high-quality data on all parts of a dataset, in the hope that results will still end up OK.

Currently, my collaboration focus is on applications in the areas of single-particle coherent diffraction imaging using X-ray free electron lasers, and modelling haplotype structure, genotype imputation, and phasing in low-coverage/high error rate genomic data. However, I am alwawys interested in pursuing and discussing other applications where statistical modelling and massive computational efforts are relevant.

Trivia: I have a history way back as a participant/medalist in international science competitions, such as IMO (mathematics), IOI (programming), IBO (biology), ACM ICPC (programming).

Are you a student seeking a thesis subject or a course project, in areas related to data analysis, HPC, or bioinformatics? Are you seeking a PhD or postdoc position? Get in touch. Projects can be tailored to a rather wide set of different backgrounds, while still staying within my research areas. We are currently eager to explore new architectures for our neural networks for genomic data (including transformers and diffusion models).


Selection of publications

Recent publications

All publications





Carl Nettelblad