Hey, I’m Cole · Open to interesting problems

Cole Agard.

I’m

I build machine learning systems and the software around them, from billion-molecule datasets to production pipelines.

  • Python ✦
  • Rust ✦
  • scikit-learn ✦
  • TensorFlow ✦
  • pandas ✦
  • NumPy ✦
  • SciPy ✦
  • Jupyter ✦
  • Plotly ✦
  • Docker ✦
  • Kubernetes ✦
  • AWS ✦
  • MLflow ✦
  • PostgreSQL ✦
  • Redis ✦
  • SQLite ✦
  • Django ✦
  • React ✦
  • Redux ✦
  • JavaScript ✦
  • Netlify ✦
  • GitHub Actions ✦

/01 About

ML engineer · data scientist · infra tinkerer

I’m a machine learning engineer and data scientist who likes turning messy data into things people actually use. I studied Computer Science at UC Berkeley, then went back for a Master of Information and Data Science.

Along the way I’ve shipped ML and software at OpenEye Scientific, Isomark Health, and Los Alamos National Laboratory: unsupervised models for drug-discovery screening, classifiers on medical device data, and a Django app used by the Department of Homeland Security.

On my own time I built Blacklight, an open-source AutoML library that evolves neural network topologies with genetic algorithms. I like distributed systems, a well-tuned pipeline, and running my own infrastructure (this site is served from a Kubernetes cluster on my desk).

molecules wrangled per dataset
1B+
molecules wrangled per dataset
saved per virtual screen
$10K
saved per virtual screen
accuracy on device illness model
94%
accuracy on device illness model
DHS users on a Django app I built
200
DHS users on a Django app I built

/02 Work

Where I’ve shipped ML and software

  1. OpenEye Scientific logo

    May ’22 – Aug ’22

    OpenEye Scientific

    Machine Learning Engineer Intern

    • Created an unsupervised ML classifier to build diverse subsets of billion-molecule datasets for pharmaceutical high-throughput virtual screening, saving upwards of $10,000 per screen.
    • Prototyped ML applications and quickly determined their viability, with documentation to support each conclusion.
    • Developed distributed approaches that scale algorithms to billions of data points with multiprocessing.
    • Python
    • Rust
    • scikit-learn
    • TensorFlow
    • pandas
    • NumPy
    • SciPy
    • Jupyter
    • Plotly
    • Docker
    • Kubernetes
    • AWS
  2. Isomark Health logo

    May ’20 – Jan ’22

    Isomark Health

    Machine Learning Engineer Intern

    • Designed and deployed a supervised classifier predicting binary indicators of animal illness from device data with 94% accuracy, and summarized results for the client.
    • Built and maintained Docker/Kubernetes data pipelines migrating diagnostic-device data to AWS S3, advancing Isomark’s cloud data practices.
    • Delivered well-documented, unit-tested systems within client-defined timeframes.
    • Python
    • TensorFlow
    • scikit-learn
    • pandas
    • NumPy
    • Jupyter
    • MLflow
    • PostgreSQL
    • Redis
    • SQLite
    • Docker
    • Kubernetes
    • AWS
  3. Los Alamos National Laboratory logo

    May ’19 – Oct ’21

    Los Alamos National Laboratory

    Software Engineering Intern

    • Built a Django application for simulating radioactive detectability, with 200 active users at the Department of Homeland Security.
    • Modernized the codebase by migrating Python 2 to Python 3, cutting the team’s technical debt.
    • Created a high-performance website to DHS specifications and delivered it ahead of deadline.
    • Python
    • Django
    • PostgreSQL
    • Redis
    • SQLite
    • Docker
    • Kubernetes
    • React

/03 Projects

Things I built because I wanted them to exist

Blacklight demo

Open-source AutoML

Blacklight

An AutoML library that uses genetic algorithms to evolve deep neural network topologies. The long-term goal is a framework that interviews a data scientist about their problem, trains a custom model with Blacklight, and deploys it behind an API.

  • Python
  • Rust
  • TensorFlow
  • scikit-learn
  • pandas
  • NumPy
  • Docker
  • Kubernetes
  • AWS
  • GitHub Actions
Proshop

Full-stack e-commerce

Proshop

A full-stack store built with Django, React, and Redux: shopping cart, reviews and ratings, search, user profiles with order history, admin product and user management, and PayPal / card checkout.

  • Python
  • Django
  • React
  • Redux
  • JavaScript
  • AWS

// Open-source contributions

/04 Education

Two Berkeley degrees and a head start

GPA 4.0
  • Statistics for Data Science Spring 2023
  • Data Science Programming Spring 2023
  • Data Science Ethics & Research Design Spring 2023
  • Applied Machine Learning Summer 2023
  • Data Engineering Summer 2023

/05 · Contact

Let’s
build something.

© 2026 Cole Agard Svelte 5 · Tailwind 4 · self-hosted on OpenShift