Hi, I'm Tyson!

Data Analyst

About

I graduated, in May 2020, from the University of Vermont (UVM) with an M.S. in mathematics. During my time at UVM — in addition to my student role — I taught undergraduate math courses to 134 students over four semesters. I also did research under my advisor, James Bagrow, in the Vermont Complex Systems Center. My thesis focused on modeling [textual] information flow in online social networks and utilized large-scale simulation, data collection, and data analysis to validate our model. I collaborated with researchers from the University of Adelaide and led our work to journal publication in Entropy .

After graduating, in the midst of the pandemic, I needed some time to re-energize and reflect before identifying a career path. During this time, I've been expanding my technical skills and developing a portfolio of practical (tackling common business problems), end-to-end projects to complement my academic work. I am joining MediaAlpha as a Data Analyst in July 2021.

I am a native Vermonter, and also obtained my B.S. in mathematics from UVM. Outside of work, my interests include: puzzles (Sudoku variants), cooking (soup & chili connoisseur), hiking, and biking.

Portfolio

I've built end-to-end data applications for: social media & technology, customer insights, and financial risk assessment. Check out a few of them below or browse more on my Github.

My spotify

in progress
  • Developed an ELT data pipeline, using Python, to analyze my daily Spotify activity.
  • Created visualizations using D3.js and built the application front-end with HTML/CSS/JavaScript.

Ice cream data dashboard

  • Assessed product performance and identified common complaints for ice cream companies — potentially increasing customer retention — by parsing customer reviews with spaCy and applying NLP methods: n-grams, sentiment analysis, and topic modeling
  • Collected all customer reviews (20,000) from four company websites using Python and Selenium for web scraping
  • Created interactive visualizations – allowing for product comparison and trend analysis – using Highcharts and R, and built a web application using R Shiny

Twitch tastes

  • Built a collaborative filtering recommendation system to suggest live streamers to Twitch.tv users, potentially increasing daily hours watched
  • Collected data on 280,000 user-streamer follows, via the Twitch API, to train our model on
  • Built a web application and RESTful API using Flask and deployed with Heroku
  • Used NetworkX to visualize and identify communities, further informing our recommendations

LendingClub loan default prediction

  • Developed a model to predict probability of default for loans on LendingClub
  • Used scikit-learn to develop, and optimize the parameters of, our feature selection and modeling pipelines
  • Used cost-sensitive metrics for imbalanced data and selected the best model from several classification models: logistic regression, KNN, random forests, XGBoost, and neural networks

Information flow in social networks (M.S. thesis)

  • Developed a novel measure and model of written information flow in online social networks
  • Estimated that a user’s text can be predicted with up to 95% accuracy using their social ties
  • Utilized a Linux cluster and Python to perform large-scale simulations (generating 100GB of data) and collect 30 million tweets to validate our proposed methodology
  • Led a team of five international researchers to publish our results

Skills

Languages

  • Python
  • R
  • MATLAB
  • SQL
  • HTML/CSS
  • JavaScript

Libraries

  • pandas
  • NumPy
  • Matplotlib/seaborn
  • scikit-learn
  • TensorFlow/Keras
  • NetworkX
  • Flask
  • R Shiny
  • D3.js
  • Highcharts

Analytical

  • Social network analysis
  • NLP
  • Recommender systems

Teaching

At UVM, I taught Fundamentals of Calculus I (MATH 019) and Applications of Finite Math (MATH 017).

My teaching page

Contact

To get in touch, send me an email at pondtyson@gmail.com or reach out on LinkedIn.