Basics of Graph Neural Networks course

Learn the basics of GNNs by joining my short (and free!) online course

Join now



Noisy labels can wreak havoc

Part 3 in the Fraud ML series

Challenges of ML with PII data

Part 2 in the Fraud ML series


An exploration of the intersection of ML and guitar circuit modeling

An Example of Graph Convolutional Networks

Controlled experiments are run on the Citeseer citation graph to understand how GCNs work

Adversarial Validation

How to use Adversarial Validation to diagnose overfitting problems

Signal : Noise

Dealing with variation in experimental results

AdaNet - Adaptive Structural Learning of Artificial Neural Networks

A paper review of Google's AdaNet AutoML technique, which learns the optimal neural network structure

Machine Learning Security

Exploring the vulnerabilities of ML models from attack by bad actors. They're more fragile than you might think

Implementing Git in Data Science

Some thoughts on how to apply version control when conducting experiments

Committing to Sanity

Why git is critical for Data Science

More than Algorithms

Some important considerations for building and deploying models in industry

Sharing your SageMaker model

How to give access to your SageMaker models to another AWS account

Overview of GANs - Part III

Learning Disentangled Representations: an introduction to InfoGAN

Overview of GANs - Part II

Improving GANs: an introduction to DC-GAN

Overview of GANs - Part I

An introduction to Generative Adversarial Networks

Bayesian Additive Regression Trees

A paper review of BART, which is the Bayesian approach to Additive Tree models

Python for data analysis

A beginner's guide to data science packages in Python

Estimation and Inference of Heterogeneous Treatment Effects using Random Forests

A paper review on using Random Forests to estimate causal impact
Zak Jost

Zak Jost

Applied Scientist @ AWS; Blogger; YouTuber

Amazon Web Services


Hello! My name is Zak and I enjoy breaking seemingly complicated things down into simple ideas. I'm self-taught in the field of ML and started writing as a way of clarifying things that I learned, and this turned out to be an enjoyable enterprise that I've continued. I offer online courses on Graph Neural Networks and release content on YouTube and this blog. And to help us all “find the others”, I am also working to build an online community. You can join us on Discord and keep up to date by subscribing to announcements here.

My day job is as an Applied Scientist at Amazon Web Services on the Graph Machine Learning team in the AI Research and Education org. Previously, I helped design and build Amazon Fraud Detector, which is a specialized AutoML solution that helps companies easily build ML-powered fraud detection systems. Prior to AWS I worked as a Data Scientist at Capital One and in the semiconductor industry as a materials scientist.

I grew up in the rural town of Troy, Missouri, got my BS/MS degrees in Physics at the University of Missouri - St. Louis, and am married with two girls. We currently reside in West Seattle on beautiful Alki Beach.


  • Machine Learning
  • Education


  • MS in Physics, 2011

    University of Missouri - St. Louis

  • BS in Physics, 2008

    University of Missouri - St. Louis



Applied Scientist

Amazon Web Services

Mar 2017 – Present Seattle
Working on the Graph ML team and building Neptune ML, which makes it easy for customers to do Machine Learning on their graph data stored in our graph database solution, Amazon Neptune. Previously, I helped design, build, and launch Amazon Fraud Detector, a fully managed service that makes it easy to build high-quality ML-powered fraud detection solutions.

Principal Data Scientist

Capital One

Sep 2015 – Feb 2017 Dallas
Helped design and build a tech stack to replace the legacy SAS systems with Python-based workflows that automated and standardized model generation and reporting.

Research Scientist / Process Engineer

SunEdison Semiconductor

Jun 2011 – May 2015 St. Louis
Specialized in ion implantation of Helium and Hydrogen for transferring nanometer scale thin films on silicon wafers in the production of Silicon-on-Insulator materials.