Jovial Design by Aaron Blythe

Random thoughts on tech, devops, meetups, conferences, and other stuff.

Data Science Portfolio

I am currently a graduate student at the University of Illinois Urbana-Champaign (the #5 Computer Science program in the nation), working on a Master’s in Computer Science with a focus on Data Science. I am maintaining a a GPA of 4.0. This page is a compilation of the work that I can share publicly. I am also working on a couple of models at work that I cannot share publicly, but can discuss privately.

Machine Learning

Deep Learning Convolutional Neural Network Paper

This was an awesome project where I worked with a research group to train a supervised Convolutional Neural Network model on 122 GB of photographs from AffectNet using PySpark, Pandas, OpenCV, Keras, NumPy, and a whole lotta cloud compute.

Check out my group’s full research paper.

Machine Learning Class

This is by far one of my favorite classes I have ever taken. We covered topics from Principal Component Analysis to Classification with Artificial Neural Networks with PyTorch and many other topics.

View my code and discussion here:

Linear Regression Class

In this class we covered a number of topics including linear regression and logistic regression.

In our group project we used the 1990 California Housing Data Set to work create a predictive model.

View my work here:

Data Visualization

Tableau Visualization of CO2 by Country

The dashboard I created shows the last 50 years of CO2 emissions per country across the world. The data is from the World Development Indicators Data.

View my work here:

D3 Dashboard using Pokémon Data

This was a fun dashboard to show off animations and interactions of elements in a D3 set of dashboards.

View my work here:

Text and Voice

Voice Controlled ChatOps Using Alexa, Hubot, and Tool API’s

I was able to create the ability to speak to my full on-call tool set using Alexa, Hubot, PagerDuty API, New Relic API and Sumologic API. It was awesome to get to present this on an IMAX screen in front of a full theater of 500 people.

Check out my presentation.

Sentiment Analysis through Natural Language Processing

In this project we mined Twitter for tweets about companies in the airline industry. We used indicators of where the tweet originated to plot the sentiment on a map. This was done in pure Python; if I were to do this again, I would do this in a Jupyter notebook as opposed to multiple Python scripts.