Author Image

August Van Hout

Using data science, anything is possible.

I am driven by challenges and teamwork, and there is no greater tool than a computer. I am an exceptionally hard-working and flexible generalist, learning and problem-solving whenever possible.

Let's get started.

Read More
Linkedin

About Me

I'm passionate about solving problems and defeating ever-greater challenges. It is the primary driver in my daily life.

Professionally, the most formative experiences have been performing machine learning engineering in a neurology lab and operations analysis in a tech manufacturing space. I learned that any person can work an incredible amount, and that big businesses provide opportunities to solve big challenges.

I first discovered Python when automating my excel work for my hourly analyst job. I was astounded by the power of computers, and I now recognize data science and analytics as the primary hard-tools for objective problem solving in my life.

For four years, I've worked with teams of diverse stakeholders to pipe data, automate reports, democratize tools, implement novel scientific methods, and use machine learning to answer questions previously unanswerable.

I am more equipped than ever to solve problems and defeat challenges.

Read More

Personal Project: Simple Neuron Field

I have a personal interest in creating simulations of neuron fields! These are models of groups of natural neurons.

I’m excited by the challenge of understanding neurons enough to create them as logical objects, and I’m further excited by the challenge of making these simulations fast enough to be practical.

I’ve even created an entire package for instantiating basic neurons in a 3D space.

Try: pip install "git+https://github.com/augustvanhout/neuron_field.git"

first service
second service

Professional Interest: Automatic Modeling

When asked to solve a modeling problem, I often find myself creating automatic modeling functions and scripts that can adapt themselves to new data.

I’ve made two so far, both using functional programming only. The script featured to the left takes any dataset, performs basic feature engineering, selects variables, and produces a full PDF report!

I’ve uploaded the script and a few example PDF reports to GitHub.

Home Prices

Home prices are arranged here by their zip code suffix.

California Standardized Tests

A Choropleth, or colormap, of standardized test pass-rates in California Counties

Facebook Status Lengths

A foundation of data science - the histogram! Displayed is the distribution of facebook status lengths in a sample from around 2010.

R-Squared

This bar graph compares how much the changes in outcome are influenced by each variable.

Diamonds

Diamond prices are plotted against their carat size and overlaid with a loess regression line. Diamonds - another data science classic.

Pearson Correlations

The highest bars in this heatmap are medical indicators most strongly correlated with kidney disease diagnosis.

Prediction vs Actuals

After making a model, we can plot our errors at different price points. If our predictions drift from the center line, our model is missing something important.

Automated Subreddit Classifier

After writing a program that learns the difference between any two subreddits, we can compare the performance with lots of subreddit pairs.

Neural Network Training

These plots show error metrics falling as a Dense Neural Network learns patterns through training.

Confusion Matrix

This graph shows whether we guessed right or wrong for a small sample of diagnoses of chronic kidney disease.

(It's a lot like "Predicted vs Actuals" above)

Score Improvement with More Features

This plot shows how adding features improves the performance of a gradient boosted trees model.

Simple Neuron Field

Thousands of simple simulated neurons are stimulated by an image on the bottom of the field, and fire in a wave.