About

I am a computational scientist passionate about using machine learning and high-performance computing to answer biological questions at scale. I have a strong foundation in probabilistic modeling, algorithm design, and software engineering, and I am experienced with analyzing large-scale omic datasets. My current research focuses on designing interpretable, multi-modal machine learning algorithms integrating genomic, transcriptomic, and proteomic blood plasma data to improve our understanding of disease and identify novel biomarkers.

Technical Skills

  • Programming languages: Python, C++, Rust, JavaScript, HTML, CSS, Workflow Description Language, Snakemake.
  • Data analysis and visualization: Pandas, Polars, scikit-learn, TensorFlow, Pyro, NetworkX, Matplotlib, d3.js.
  • Molecular data analysis: Illumina (genomic, metagenomic, transcriptomic), PacBio (genomic), Olink (proteomics).
  • Software: Linux, Git, Google Cloud Platform, PostgreSQL, Adobe Illustrator.

Professional experience

Broad Institute of MIT and Harvard, Cambridge, MA

  • Computational Scientist II, Data Sciences Platform, Sep 2024 - Current
    • Enabled the characterization of structurally diverse genomic regions in Plasmodium falciparum by redesigning a hidden markov model for recombination-aware sequence alignment, resulting in 360x faster inference.
    • Developed a data pre-processing and quality control pipeline, curating disease phenotypes from electronic health records for 10,000 participants in a national biobank and assessing the quality of associated genomic, transcriptomic, and proteomic samples.
    • Working on designing interpretable, multi-modal machine learning models to cluster biobank participants, integrating genomic, transcriptomic, and proteomic blood plasma data to improve our understanding of disease and identify novel biomarkers.
GitHub - castcollab/tesserae2: Tesserae2: Fast recombination-aware global and local alignment.
Tesserae2: Fast recombination-aware global and local alignment. - castcollab/tesserae2
GitHubcastcollab
  • Computational Associate II, Bacterial Genomics Lab, Nov 2017 - Aug 2024
    • Designed and implemented a new optimal partial order alignment algorithm in Rust, which was, on average, 4.1x faster than similar tools, enabling DNA sequence-to-graph alignments previously impossible.
    • Led the development of Python and C++-based software specifically designed to track low-abundance (>0.1%) bacterial strains in complex microbial communities (e.g., the human gut microbiome) using whole metagenome sequencing data.
    • Developed a cloud-based pipeline using the Workflow Description Language (WDL) and Docker to enable fast and reproducible characterization of thousands of metagenomic samples.
    • Obtained detailed insight into the E. coli strain-level dynamics in a year-long longitudinal microbiome study of women with recurrent urinary tract infections (UTIs), revealing unexpected similarities with a healthy control group and that the UTI-causing strain is rarely cleared from the gut after antibiotics.
StrainGE: Strain Genome Explorer
StrainGE is a toolkit for tracking and characterizing low-abundance strains in complex microbial communities. It enables detailed insights into the bacterial strain-level diversity of whole metagenomic sequencing samples.
Lucas van DijkLucas van Dijk
Gut microbiome dysbiosis linked with recurrent UTIs
More than half of the women in the US get a urinary tract infection (UTI) in their lifetime, which frequently becomes recurrent. In this paper, we investigated the role of the gut microbiome in facilitating recurrence.
Lucas van DijkLucas van Dijk

DSM-firmenich, Delft, The Netherlands

  • Intern, Jun 2016 - Oct 2016
    • Reviewed literature and proposed a plan to analyze a large-scale protein production problem with Bacillus subtilis using genome-scale metabolic models.

Studio bereikbaar, Rotterdam, The Netherlands

  • Software Engineer, May 2013 - Jun 2016
    • Led the development of web-based geographic information systems (GIS) tools using Django and PostgreSQL, facilitating collaborative project planning and design between engineering colleagues.
    • These tools helped win several multi-million euro infrastructure construction contracts with the government.

Thales, Delft, The Netherlands

  • Intern, Sep 2012 - Nov 2012
    • Developed a GPU-accelerated tool to analyze the radar reflectivity of navy ships using nVidia's OptiX CUDA ray tracing library.

Extracurricular Projects

Python Software Foundation (VisPy Google Summer of Code 2015)

  • Implemented a high-performance graph visualization system in Python and OpenGL, including several automatic graph layout algorithms.
  • Contributed the open-source code upstream to VisPy, a high-performance scientific data visualization software
    library.
Drawing arbitrary shapes with OpenGL points
Part of my Google Summer of Code project involves porting several arrow heads from Glumpy to Vispy. I also want to make a slight change to them: the arrow heads in Glumpy include an arrow body, I want to remove that to make sure you can put an arrow head
Lucas van DijkLucas van Dijk
Home — VisPy
VisPy - Home

Delft University of Technology, "Helios 3D team"

  • Built a large 3x3x1 meter 3D RGB LED cube display as part of a student team.
  • Architected and implemented the C++ embedded software responsible for receiving the image to display over Wi-Fi and driving individual LEDs to the correct color.

Education

  • PhD Bioinformatics. Delft University of Technology. 2025.
    • PhD in the bacterial genomics lab at the Broad Institute in collaboration with the bioinformatics lab at TU Delft.
  • MSc Computer Science. Delft University of Technology. 2017.
  • BEng Electrical Engineering. The Hague University of Applied Sciences. 2013.