Alex Wurts

Programming

I am a Computer Science Major and Data Science at Worcester Polytechnic Institute. Throughout my time in college, I have explored many areas from Deep Learning to Front End development. I am always looking for new areas to explore and learn.

Flying

I am a Private Pilot. I trained at Hanscom Airport in Bedford MA flying Piper Warrior IIIs. I eventually want to buy or build a small bush plane and explore the US.

Programming Tutorials

Knowing how to program can change someone's life. Learning how to program can be incredibly challenging, and I know my knowledge could help someone learn faster. I try to do what I can to make learning programming more achievable. I find after I teach a subject I understand it much better.

I create video tutorials and articles to help pass on my knowledge to allow as many people as possible to learn the awesome world of programming.

Senior Thesis Intern August 2019 - October 2019

MIT Lincoln Laboratory and WPI Lexington, MA

Tauopathies are hypothesized to be a primary cause of Alzheimer's disease. Labeling tau proteins in a brain image is a task requiring high expertise and significant time. My project successfully developed a Deep Learning solution to segment and quantify important cells in brain scans.

Built using the SegNet Convolutional Neural Network architecture
Final model saves Doctors many hours of labeling and enables improved diagnosis of the disease.
Wrote parallelized code on the MIT supercomputer to process and stitch together 10 billion pixel brain scans in under 30 minutes.

Technology: Python, Pandas, Pillow, Numpy, PyTorch, Sklearn, Slurm, MIT LL Super Computer, Jupyter Notebooks, Anaconda, Convolutional Neural Networks, Matplotlib

Deep Learning Intern May 2019 - August 2019

MIT Lincoln Laboratory Lexington, MA

Implemented state of art deep neural networks for the person re-identification in a camera surveillance system problem.
Used a supercomputer to train and evaluate many data set and model configurations.
Developed innovative image labeling application using deep neural networks
Reduced time to label a dataset by 5 times, saving days worth of time in future deployments.
Final model reduced classification time by 2x over non neural network approach.

Technology: Python 3, Pandas, Pillow, Numpy, PyTorch, Sklearn, Slurm, MIT LL Super Computer, Jupyter Notebooks

Software Engineer Intern Jun 2018 - Aug 2018

Olympus Waltham, MA

Designed and developed a website to visualize metrics produced by a company product
Implemented an interface that allows employees to set rules that when triggered send notifications about impending failure, so they can perform preventative maintenance.
Saved thousands of dollars by creating a robust system to understand device alerts.
Created a REST API for the backend, to provide a portable solution that could eventually be used across many platforms.

Technology: Javascript, React, Node, MongoDB, Microsoft Azure, Material UI, Express

Pilot Weather Visualizations PilotWeatherVis.com Code - Github

Role:Creator

I created a user interface to help training pilots understand confusing weather forecasting systems such as METARs (weather now), and TAFs (weather soon). Visualizations created with weather data searchable for any airport in the US. Data sourced from the Aviation Weather Center.

Skills: React, Node, JS, REST API, Express, D3js, NWS ADDS API, Github

Neural Network Implementation Code - Github

Role: Creator

Implemented a neural network from scratch that trains on the MNIST data set. The network architecture has two hidden layers composed of an adjustable amount of hidden neurons. Created a hyperparameter optimization algorithm that optimizes each hyperparameter individually and combines them together.

Skills: Python, Numpy, Matplotlib, Scikit-Learn

Visualization of Gradient Descent Training

Best Routes BestRoutes.io

Role: Creator

Developed a full-stack web app to create a user-sourced database of the best rides and runs in a certain area. Utilizes a React, Express, Node, MongoDB Stack. Retrieves data automatically from Strava API and Google API. Code available upon request.

Skills: Reactjs, Nodejs, Expressjs, MongoDB, Google Cloud Platform, Leaflet Maps, D3js, Google Maps API, Strava API, Material UI

Resumace Resumace

Role: Creator

When applying to new jobs it is time-consuming to customize a resume for the specific job. When completed my application will allow a user to create a new resume in as little as 30 seconds. Additionally, formatting resumes can be difficult and prone to errors. The application formats the data automatically and allows the user to select different styles. Utilizes a React, Express, Node, MySQL database Stack.

Skills: React, Node, Express, MySQL, Google Cloud Platform, Square Payments, HTML, CSS, Javascript

Android Solitaire Google Play Store Code - Github

Role: Creator

Developed Solitaire to create a version with no ads. Released on Google Play. Follow the link for more details.

Skills: Java, Android Studio, Canvas, Google Play Store

CIRA: Chromosome Image Recognition Algorithm Code - Github

Role: Group Member

Developed a machine learning algorithm to predict the health of new cells. Used K-Means to cluster all known cells into 2 groups, and then new cells were labeled according to the nearest group. The Accuracy was 79.1%. Multi-Level Perceptron attained an accuracy of 86.025% and Logistic Regression was 87%.

Skills: Python, Numpy, Matplotlib, Scikit-learn, Scipy, Anaconda, Pillow

Example Clustering of Cells.

RideVis: An Improved Strava Visualization Ride Vis Code - Github

Role: Created for Data Vis Class

Re-made the Strava Profile Visualization. Aggregates my Strava ride data using python into varying time chunks. Then uses D3js to plot them and make it interactive. Click on the column to zoom into that period, and click again to pull up the Strava ride.

Skills: Python, Numpy, Pandas, Javascript, D3, Github

Overarching Project Goals