multidms
A JAX-based Python package for jointly modeling multiple Deep Mutational Scanning datasets. Provides tools for data preparation, model fitting, out-of-sample prediction, and interactive visualization of parameter values.
About
I am an Engineer with a background in Computer & Information Science and a focus on research in the Biological Sciences. I have seven years of hands-on experience conducting various flavors of methods development, implementation, and analysis in an academic research environment.
In practice, my work has consisted primarily of creating machine learning-based software tools that allow researchers to analyze and interpret complex biological datasets. My research often involves using cutting-edge statistical methods to filter artifacts, normalize/encode data, fit models, and visualize data so that results are scientifically sound and practically useful.
I take pride in clearly communicating my work and enjoy the discussions that follow.
Skills
Languages
- Python
- R
- C/C++
- Bash
- Java
- JavaScript
ML & Data Science
- JAX
- TensorFlow/Keras
- PyTorch
- Scikit-learn
- Pandas
- Polars
- NumPy
- Xarray
- SciPy
- BioPython
Visualization
- Matplotlib
- Plotnine
- Altair
Tools & Infrastructure
- Git
- GitHub Actions
- Docker
- Apptainer
- Conda/Mamba
- Snakemake
- Nextflow
- SLURM HPC
- AWS S3
- LaTeX
Mathematics
- Bayesian Statistics
- Linear Algebra
- Discrete Math
- Calculus
Projects
phippery
A software suite for analyzing phage immunoprecipitation sequencing (PhIP-Seq) data. Includes a Nextflow pipeline for processing, a Python API for normalization and enrichment analysis, and a Streamlit app for interactive visualization.
ReLERNN
Recombination Landscape Estimation using Recurrent Neural Networks. A deep learning method for estimating genome-wide recombination maps that is accurate even with small numbers of pooled or individually sequenced genomes.
Experience
Bioinformatics Specialist
Howard Hughes Medical Institute
Developing novel approaches for jointly modeling deep mutational scanning datasets using JAX. Building Nextflow pipelines for studying affinity maturation in germinal centers, managing sensitive data on HPC and AWS, and producing publication-quality figures.
Bioinformatics Analyst II
Fred Hutchinson Cancer Center
Designed algorithms for mutational fitness estimation across 13M+ viral genomes. Built and deployed public Nextstrain pages for genomic surveillance of SARS-CoV-2, Influenza, and Lassa virus. Mentored interns, students, and postdocs in computational methods.
Research Associate
Fred Hutchinson Cancer Center
Developed the phippery software suite for PhIP-Seq data analysis, including a Nextflow pipeline, Python API, and Streamlit visualization app. Contributed to global epistasis modeling with PyTorch. Led cross-lab team meetings.
Graduate Teaching Fellow
University of Oregon
Taught graduate-level Python/Bash scripting and algorithmic thinking. Built Docker containers and a Jupyter Notebook auto-grading pipeline. Led discussions and graded ~100 assignments per week.
Scientific Programmer
University of Oregon
Applied software engineering best practices to open-source tools including tskit, stdpopsim, and ReLERNN. Evaluated deep learning models for recombination rate estimation. Built Snakemake pipelines for population genetics analyses.
Undergraduate Research Assistant
University of Oregon
Implemented tree sequence recording in the SLiM population genetics simulator. Benchmarked large-scale simulations for an undergraduate honors thesis. Developed and analyzed evolutionary models of stickleback fish migration.